id	sid	tid	token	lemma	pos
easat-9471	1	1	edelweiss	edelweiss	PROPN
easat-9471	1	2	applied	apply	VERB
easat-9471	1	3	science	science	NOUN
easat-9471	1	4	and	and	CCONJ
easat-9471	1	5	technology	technology	NOUN
easat-9471	1	6	issn	issn	PROPN
easat-9471	1	7	:	:	PUNCT
easat-9471	1	8	2576	2576	NUM
easat-9471	1	9	-	-	SYM
easat-9471	1	10	8484	8484	NUM
easat-9471	1	11	vol	vol	NOUN
easat-9471	1	12	.	.	PROPN
easat-9471	2	1	9	9	NUM
easat-9471	2	2	,	,	PUNCT
easat-9471	2	3	no	no	INTJ
easat-9471	2	4	.	.	NOUN
easat-9471	2	5	8	8	NUM
easat-9471	2	6	,	,	PUNCT
easat-9471	2	7	871	871	NUM
easat-9471	2	8	-	-	SYM
easat-9471	2	9	881	881	NUM
easat-9471	2	10	2025	2025	NUM
easat-9471	2	11	publisher	publisher	NOUN
easat-9471	2	12	:	:	PUNCT
easat-9471	2	13	learning	learn	VERB
easat-9471	2	14	gate	gate	NOUN
easat-9471	2	15	doi	doi	PROPN
easat-9471	2	16	:	:	PUNCT
easat-9471	2	17	10.55214/2576	10.55214/2576	NUM
easat-9471	2	18	-	-	PUNCT
easat-9471	2	19	8484.v9i8.9471	8484.v9i8.9471	NOUN
easat-9471	3	1	©	©	PROPN
easat-9471	3	2	2025	2025	NUM
easat-9471	3	3	by	by	ADP
easat-9471	3	4	the	the	DET
easat-9471	3	5	authors	author	NOUN
easat-9471	3	6	;	;	PUNCT
easat-9471	3	7	licensee	licensee	PROPN
easat-9471	3	8	learning	learning	NOUN
easat-9471	3	9	gate	gate	NOUN
easat-9471	3	10	©	©	PROPN
easat-9471	3	11	2025	2025	NUM
easat-9471	3	12	by	by	ADP
easat-9471	3	13	the	the	DET
easat-9471	3	14	authors	author	NOUN
easat-9471	3	15	;	;	PUNCT
easat-9471	3	16	licensee	licensee	PROPN
easat-9471	3	17	learning	learning	NOUN
easat-9471	3	18	gate	gate	NOUN
easat-9471	3	19	history	history	NOUN
easat-9471	3	20	:	:	PUNCT
easat-9471	3	21	received	receive	VERB
easat-9471	3	22	:	:	PUNCT
easat-9471	3	23	3	3	NUM
easat-9471	3	24	june	june	PROPN
easat-9471	3	25	2025	2025	NUM
easat-9471	3	26	;	;	PUNCT
easat-9471	3	27	revised	revise	VERB
easat-9471	3	28	:	:	PUNCT
easat-9471	3	29	18	18	NUM
easat-9471	3	30	june	june	PROPN
easat-9471	3	31	2025	2025	NUM
easat-9471	3	32	;	;	PUNCT
easat-9471	3	33	accepted	accept	VERB
easat-9471	3	34	:	:	PUNCT
easat-9471	3	35	23	23	NUM
easat-9471	3	36	june	june	PROPN
easat-9471	3	37	2025	2025	NUM
easat-9471	3	38	;	;	PUNCT
easat-9471	3	39	published	publish	VERB
easat-9471	3	40	:	:	PUNCT
easat-9471	3	41	15	15	NUM
easat-9471	3	42	august	august	PROPN
easat-9471	3	43	2025	2025	NUM
easat-9471	3	44	*	*	PUNCT
easat-9471	3	45	correspondence	correspondence	NOUN
easat-9471	3	46	:	:	PUNCT
easat-9471	3	47	kwonhj@seoultech.ac.kr	kwonhj@seoultech.ac.kr	NOUN
easat-9471	3	48	optimizing	optimize	VERB
easat-9471	3	49	collaborative	collaborative	ADJ
easat-9471	3	50	filtering	filtering	NOUN
easat-9471	3	51	recommender	recommender	NOUN
easat-9471	3	52	systems	system	NOUN
easat-9471	3	53	with	with	ADP
easat-9471	3	54	the	the	DET
easat-9471	3	55	grpo	grpo	NOUN
easat-9471	3	56	reinforcement	reinforcement	NOUN
easat-9471	3	57	learning	learning	NOUN
easat-9471	3	58	algorithm	algorithm	PROPN
easat-9471	3	59	hwi	hwi	PROPN
easat-9471	3	60	jeon1	jeon1	PROPN
easat-9471	3	61	,	,	PUNCT
easat-9471	3	62	chan	chan	PROPN
easat-9471	3	63	-	-	PUNCT
easat-9471	3	64	ho	ho	PROPN
easat-9471	3	65	lee2	lee2	PROPN
easat-9471	3	66	,	,	PUNCT
easat-9471	3	67	jong	jong	PROPN
easat-9471	3	68	-	-	PUNCT
easat-9471	3	69	geun	geun	PROPN
easat-9471	3	70	choi3	choi3	PROPN
easat-9471	3	71	,	,	PUNCT
easat-9471	3	72	hyuk	hyuk	NOUN
easat-9471	3	73	-	-	PUNCT
easat-9471	3	74	jin	jin	NOUN
easat-9471	3	75	kwon4	kwon4	PROPN
easat-9471	3	76	*	*	NOUN
easat-9471	3	77	1,2,3,4department	1,2,3,4department	NUM
easat-9471	3	78	of	of	ADP
easat-9471	3	79	defense	defense	NOUN
easat-9471	3	80	ai	ai	PROPN
easat-9471	3	81	applications	application	NOUN
easat-9471	3	82	,	,	PUNCT
easat-9471	3	83	seoul	seoul	PROPN
easat-9471	3	84	national	national	PROPN
easat-9471	3	85	university	university	PROPN
easat-9471	3	86	of	of	ADP
easat-9471	3	87	science	science	NOUN
easat-9471	3	88	and	and	CCONJ
easat-9471	3	89	technology	technology	NOUN
easat-9471	3	90	,	,	PUNCT
easat-9471	3	91	seoul	seoul	PROPN
easat-9471	3	92	01811	01811	NUM
easat-9471	3	93	,	,	PUNCT
easat-9471	3	94	republic	republic	NOUN
easat-9471	3	95	of	of	ADP
easat-9471	3	96	korea	korea	PROPN
easat-9471	3	97	;	;	PUNCT
easat-9471	3	98	hwi042595@seoultech.ac.kr	hwi042595@seoultech.ac.kr	PROPN
easat-9471	3	99	(	(	PUNCT
easat-9471	3	100	h.j	h.j	PROPN
easat-9471	3	101	.	.	PROPN
easat-9471	3	102	)	)	PUNCT
easat-9471	4	1	cksgh1211@seoultech.ac.kr	cksgh1211@seoultech.ac.kr	PROPN
easat-9471	4	2	(	(	PUNCT
easat-9471	4	3	c.h.l	c.h.l	NOUN
easat-9471	4	4	.	.	PUNCT
easat-9471	4	5	)	)	PUNCT
easat-9471	4	6	choijon1@seoultech.ac.kr	choijon1@seoultech.ac.kr	X
easat-9471	4	7	(	(	PUNCT
easat-9471	4	8	j.g.c	j.g.c	PROPN
easat-9471	4	9	.	.	PUNCT
easat-9471	4	10	)	)	PUNCT
easat-9471	5	1	kwonhj@seoultech.ac.kr	kwonhj@seoultech.ac.kr	NOUN
easat-9471	5	2	(	(	PUNCT
easat-9471	5	3	h.j.k	h.j.k	NOUN
easat-9471	5	4	.	.	PUNCT
easat-9471	5	5	)	)	PUNCT
easat-9471	5	6	.	.	PUNCT
easat-9471	6	1	abstract	abstract	ADJ
easat-9471	6	2	:	:	PUNCT
easat-9471	6	3	collaborative	collaborative	ADJ
easat-9471	6	4	filtering	filtering	NOUN
easat-9471	6	5	recommender	recommender	NOUN
easat-9471	6	6	systems	system	NOUN
easat-9471	6	7	primarily	primarily	ADV
easat-9471	6	8	focus	focus	VERB
easat-9471	6	9	on	on	ADP
easat-9471	6	10	short	short	ADJ
easat-9471	6	11	-	-	PUNCT
easat-9471	6	12	term	term	NOUN
easat-9471	6	13	prediction	prediction	NOUN
easat-9471	6	14	accuracy	accuracy	NOUN
easat-9471	6	15	but	but	CCONJ
easat-9471	6	16	exhibit	exhibit	NOUN
easat-9471	6	17	limitations	limitation	NOUN
easat-9471	6	18	concerning	concern	VERB
easat-9471	6	19	long	long	ADJ
easat-9471	6	20	-	-	PUNCT
easat-9471	6	21	term	term	NOUN
easat-9471	6	22	user	user	NOUN
easat-9471	6	23	satisfaction	satisfaction	NOUN
easat-9471	6	24	and	and	CCONJ
easat-9471	6	25	content	content	NOUN
easat-9471	6	26	diversity	diversity	NOUN
easat-9471	6	27	.	.	PUNCT
easat-9471	7	1	in	in	ADP
easat-9471	7	2	this	this	DET
easat-9471	7	3	paper	paper	NOUN
easat-9471	7	4	,	,	PUNCT
easat-9471	7	5	we	we	PRON
easat-9471	7	6	reinterpret	reinterpret	VERB
easat-9471	7	7	user	user	NOUN
easat-9471	7	8	-	-	PUNCT
easat-9471	7	9	item	item	NOUN
easat-9471	7	10	interaction	interaction	NOUN
easat-9471	7	11	data	datum	NOUN
easat-9471	7	12	as	as	ADP
easat-9471	7	13	reinforcement	reinforcement	NOUN
easat-9471	7	14	learning	learning	NOUN
easat-9471	7	15	with	with	ADP
easat-9471	7	16	verifiable	verifiable	ADJ
easat-9471	7	17	rewards	reward	NOUN
easat-9471	7	18	and	and	CCONJ
easat-9471	7	19	introduce	introduce	VERB
easat-9471	7	20	the	the	DET
easat-9471	7	21	group	group	NOUN
easat-9471	7	22	relative	relative	ADJ
easat-9471	7	23	policy	policy	NOUN
easat-9471	7	24	optimization	optimization	NOUN
easat-9471	7	25	(	(	PUNCT
easat-9471	7	26	grpo	grpo	NOUN
easat-9471	7	27	)	)	PUNCT
easat-9471	7	28	reinforcement	reinforcement	NOUN
easat-9471	7	29	learning	learning	NOUN
easat-9471	7	30	algorithm	algorithm	NOUN
easat-9471	7	31	,	,	PUNCT
easat-9471	7	32	originally	originally	ADV
easat-9471	7	33	proposed	propose	VERB
easat-9471	7	34	in	in	ADP
easat-9471	7	35	the	the	DET
easat-9471	7	36	large	large	ADJ
easat-9471	7	37	language	language	NOUN
easat-9471	7	38	model	model	NOUN
easat-9471	7	39	domain	domain	NOUN
easat-9471	7	40	,	,	PUNCT
easat-9471	7	41	to	to	ADP
easat-9471	7	42	collaborative	collaborative	ADJ
easat-9471	7	43	filtering	filtering	NOUN
easat-9471	7	44	model	model	NOUN
easat-9471	7	45	fine	fine	ADV
easat-9471	7	46	-	-	PUNCT
easat-9471	7	47	tuning	tuning	NOUN
easat-9471	7	48	for	for	ADP
easat-9471	7	49	the	the	DET
easat-9471	7	50	first	first	ADJ
easat-9471	7	51	time	time	NOUN
easat-9471	7	52	.	.	PUNCT
easat-9471	8	1	grpo	grpo	NOUN
easat-9471	8	2	directly	directly	ADV
easat-9471	8	3	updates	update	VERB
easat-9471	8	4	policies	policy	NOUN
easat-9471	8	5	without	without	ADP
easat-9471	8	6	separate	separate	ADJ
easat-9471	8	7	critic	critic	NOUN
easat-9471	8	8	networks	network	NOUN
easat-9471	8	9	,	,	PUNCT
easat-9471	8	10	balancing	balance	VERB
easat-9471	8	11	exploration	exploration	NOUN
easat-9471	8	12	and	and	CCONJ
easat-9471	8	13	exploitation	exploitation	NOUN
easat-9471	8	14	while	while	SCONJ
easat-9471	8	15	optimizing	optimize	VERB
easat-9471	8	16	long	long	ADJ
easat-9471	8	17	-	-	PUNCT
easat-9471	8	18	term	term	NOUN
easat-9471	8	19	user	user	NOUN
easat-9471	8	20	engagement	engagement	NOUN
easat-9471	8	21	.	.	PUNCT
easat-9471	9	1	in	in	ADP
easat-9471	9	2	experiments	experiment	NOUN
easat-9471	9	3	conducted	conduct	VERB
easat-9471	9	4	on	on	ADP
easat-9471	9	5	amazon	amazon	NOUN
easat-9471	9	6	review	review	VERB
easat-9471	9	7	datasets	dataset	NOUN
easat-9471	9	8	covering	cover	VERB
easat-9471	9	9	baby	baby	NOUN
easat-9471	9	10	products	product	NOUN
easat-9471	9	11	,	,	PUNCT
easat-9471	9	12	video	video	NOUN
easat-9471	9	13	games	game	NOUN
easat-9471	9	14	,	,	PUNCT
easat-9471	9	15	and	and	CCONJ
easat-9471	9	16	industrial	industrial	ADJ
easat-9471	9	17	&	&	CCONJ
easat-9471	9	18	scientific	scientific	ADJ
easat-9471	9	19	categories	category	NOUN
easat-9471	9	20	,	,	PUNCT
easat-9471	9	21	the	the	DET
easat-9471	9	22	grpooptimized	grpooptimize	VERB
easat-9471	9	23	model	model	NOUN
easat-9471	9	24	achieved	achieve	VERB
easat-9471	9	25	up	up	ADP
easat-9471	9	26	to	to	ADP
easat-9471	9	27	15.16	15.16	NUM
easat-9471	9	28	%	%	NOUN
easat-9471	9	29	improvement	improvement	NOUN
easat-9471	9	30	in	in	ADP
easat-9471	9	31	recall@10	recall@10	NOUN
easat-9471	9	32	compared	compare	VERB
easat-9471	9	33	to	to	ADP
easat-9471	9	34	baseline	baseline	NOUN
easat-9471	9	35	models	model	NOUN
easat-9471	9	36	.	.	PUNCT
easat-9471	10	1	additionally	additionally	ADV
easat-9471	10	2	,	,	PUNCT
easat-9471	10	3	we	we	PRON
easat-9471	10	4	revealed	reveal	VERB
easat-9471	10	5	that	that	SCONJ
easat-9471	10	6	user	user	NOUN
easat-9471	10	7	embeddings	embedding	NOUN
easat-9471	10	8	from	from	ADP
easat-9471	10	9	graph	graph	NOUN
easat-9471	10	10	-	-	PUNCT
easat-9471	10	11	based	base	VERB
easat-9471	10	12	collaborative	collaborative	ADJ
easat-9471	10	13	filtering	filtering	NOUN
easat-9471	10	14	architectures	architecture	NOUN
easat-9471	10	15	positively	positively	ADV
easat-9471	10	16	contribute	contribute	VERB
easat-9471	10	17	to	to	ADP
easat-9471	10	18	grpo	grpo	NOUN
easat-9471	10	19	algorithm	algorithm	NOUN
easat-9471	10	20	optimization	optimization	NOUN
easat-9471	10	21	,	,	PUNCT
easat-9471	10	22	whereas	whereas	SCONJ
easat-9471	10	23	positional	positional	ADJ
easat-9471	10	24	embeddings	embedding	NOUN
easat-9471	10	25	from	from	ADP
easat-9471	10	26	sequential	sequential	ADJ
easat-9471	10	27	collaborative	collaborative	ADJ
easat-9471	10	28	filtering	filter	VERB
easat-9471	10	29	architectures	architecture	NOUN
easat-9471	10	30	impede	impede	VERB
easat-9471	10	31	optimization	optimization	NOUN
easat-9471	10	32	performance	performance	NOUN
easat-9471	10	33	.	.	PUNCT
easat-9471	11	1	these	these	DET
easat-9471	11	2	findings	finding	NOUN
easat-9471	11	3	empirically	empirically	ADV
easat-9471	11	4	validate	validate	VERB
easat-9471	11	5	the	the	DET
easat-9471	11	6	effectiveness	effectiveness	NOUN
easat-9471	11	7	of	of	ADP
easat-9471	11	8	the	the	DET
easat-9471	11	9	grpo	grpo	NOUN
easat-9471	11	10	algorithm	algorithm	NOUN
easat-9471	11	11	as	as	ADP
easat-9471	11	12	a	a	DET
easat-9471	11	13	robust	robust	ADJ
easat-9471	11	14	approach	approach	NOUN
easat-9471	11	15	for	for	ADP
easat-9471	11	16	recommender	recommender	NOUN
easat-9471	11	17	system	system	NOUN
easat-9471	11	18	model	model	NOUN
easat-9471	11	19	optimization	optimization	NOUN
easat-9471	11	20	.	.	PUNCT
easat-9471	12	1	keywords	keyword	NOUN
easat-9471	12	2	:	:	PUNCT
easat-9471	12	3	collaborative	collaborative	ADJ
easat-9471	12	4	filtering	filtering	NOUN
easat-9471	12	5	,	,	PUNCT
easat-9471	12	6	grpo	grpo	NOUN
easat-9471	12	7	,	,	PUNCT
easat-9471	12	8	recommender	recommender	NOUN
easat-9471	12	9	system	system	NOUN
easat-9471	12	10	,	,	PUNCT
easat-9471	12	11	reinforcement	reinforcement	NOUN
easat-9471	12	12	learning	learning	NOUN
easat-9471	12	13	,	,	PUNCT
easat-9471	12	14	rlvr	rlvr	NOUN
easat-9471	12	15	.	.	PUNCT
easat-9471	13	1	1	1	X
easat-9471	13	2	.	.	X
easat-9471	13	3	introduction	introduction	NOUN
easat-9471	13	4	recommender	recommender	NOUN
easat-9471	13	5	systems	system	NOUN
easat-9471	13	6	are	be	AUX
easat-9471	13	7	core	core	NOUN
easat-9471	13	8	technologies	technology	NOUN
easat-9471	13	9	used	use	VERB
easat-9471	13	10	in	in	ADP
easat-9471	13	11	various	various	ADJ
easat-9471	13	12	commercial	commercial	ADJ
easat-9471	13	13	platforms	platform	NOUN
easat-9471	13	14	,	,	PUNCT
easat-9471	13	15	such	such	ADJ
easat-9471	13	16	as	as	ADP
easat-9471	13	17	ecommerce	ecommerce	NOUN
easat-9471	13	18	and	and	CCONJ
easat-9471	13	19	streaming	streaming	NOUN
easat-9471	13	20	services	service	NOUN
easat-9471	13	21	,	,	PUNCT
easat-9471	13	22	to	to	PART
easat-9471	13	23	support	support	VERB
easat-9471	13	24	decision	decision	NOUN
easat-9471	13	25	-	-	PUNCT
easat-9471	13	26	making	making	NOUN
easat-9471	13	27	by	by	ADP
easat-9471	13	28	providing	provide	VERB
easat-9471	13	29	personalized	personalized	ADJ
easat-9471	13	30	content	content	NOUN
easat-9471	13	31	based	base	VERB
easat-9471	13	32	on	on	ADP
easat-9471	13	33	users	user	NOUN
easat-9471	13	34	'	'	PART
easat-9471	13	35	past	past	ADJ
easat-9471	13	36	behavior	behavior	NOUN
easat-9471	13	37	.	.	PUNCT
easat-9471	14	1	recommender	recommender	NOUN
easat-9471	14	2	systems	system	NOUN
easat-9471	14	3	are	be	AUX
easat-9471	14	4	classified	classify	VERB
easat-9471	14	5	into	into	ADP
easat-9471	14	6	collaborative	collaborative	ADJ
easat-9471	14	7	filtering	filtering	NOUN
easat-9471	14	8	,	,	PUNCT
easat-9471	14	9	content	content	NOUN
easat-9471	14	10	-	-	PUNCT
easat-9471	14	11	based	base	VERB
easat-9471	14	12	filtering	filtering	NOUN
easat-9471	14	13	,	,	PUNCT
easat-9471	14	14	and	and	CCONJ
easat-9471	14	15	hybrid	hybrid	ADJ
easat-9471	14	16	methods	method	NOUN
easat-9471	14	17	that	that	PRON
easat-9471	14	18	integrate	integrate	VERB
easat-9471	14	19	the	the	DET
easat-9471	14	20	two	two	NUM
easat-9471	14	21	.	.	PUNCT
easat-9471	15	1	collaborative	collaborative	ADJ
easat-9471	15	2	filtering	filter	VERB
easat-9471	15	3	methods	method	NOUN
easat-9471	15	4	are	be	AUX
easat-9471	15	5	rely	rely	ADJ
easat-9471	15	6	on	on	ADP
easat-9471	15	7	the	the	DET
easat-9471	15	8	theoretical	theoretical	ADJ
easat-9471	15	9	assumption	assumption	NOUN
easat-9471	15	10	that	that	SCONJ
easat-9471	15	11	users	user	NOUN
easat-9471	15	12	exhibiting	exhibit	VERB
easat-9471	15	13	similar	similar	ADJ
easat-9471	15	14	behaviors	behavior	NOUN
easat-9471	15	15	share	share	VERB
easat-9471	15	16	similar	similar	ADJ
easat-9471	15	17	preferences	preference	NOUN
easat-9471	15	18	.	.	PUNCT
easat-9471	16	1	collaborative	collaborative	ADJ
easat-9471	16	2	filtering	filter	VERB
easat-9471	16	3	methods	method	NOUN
easat-9471	16	4	are	be	AUX
easat-9471	16	5	further	far	ADV
easat-9471	16	6	divided	divide	VERB
easat-9471	16	7	into	into	ADP
easat-9471	16	8	sequential	sequential	ADJ
easat-9471	16	9	collaborative	collaborative	ADJ
easat-9471	16	10	filtering	filtering	NOUN
easat-9471	16	11	methods	method	NOUN
easat-9471	16	12	and	and	CCONJ
easat-9471	16	13	graph	graph	NOUN
easat-9471	16	14	-	-	PUNCT
easat-9471	16	15	based	base	VERB
easat-9471	16	16	collaborative	collaborative	ADJ
easat-9471	16	17	filtering	filtering	NOUN
easat-9471	16	18	methods	method	NOUN
easat-9471	16	19	,	,	PUNCT
easat-9471	16	20	with	with	ADP
easat-9471	16	21	the	the	DET
easat-9471	16	22	former	former	ADJ
easat-9471	16	23	learning	learn	VERB
easat-9471	16	24	users	user	NOUN
easat-9471	16	25	'	'	PART
easat-9471	16	26	temporal	temporal	ADJ
easat-9471	16	27	interaction	interaction	NOUN
easat-9471	16	28	relationships	relationship	NOUN
easat-9471	16	29	and	and	CCONJ
easat-9471	16	30	the	the	DET
easat-9471	16	31	latter	latter	ADJ
easat-9471	16	32	learning	learn	VERB
easat-9471	16	33	structural	structural	ADJ
easat-9471	16	34	relationships	relationship	NOUN
easat-9471	16	35	between	between	ADP
easat-9471	16	36	users	user	NOUN
easat-9471	16	37	and	and	CCONJ
easat-9471	16	38	items	item	NOUN
easat-9471	16	39	.	.	PUNCT
easat-9471	17	1	representative	representative	ADJ
easat-9471	17	2	sequential	sequential	ADJ
easat-9471	17	3	collaborative	collaborative	ADJ
easat-9471	17	4	filtering	filtering	NOUN
easat-9471	17	5	methods	method	NOUN
easat-9471	17	6	include	include	VERB
easat-9471	17	7	fpmc	fpmc	PROPN
easat-9471	17	8	[	[	X
easat-9471	17	9	1	1	NUM
easat-9471	17	10	]	]	PUNCT
easat-9471	17	11	which	which	PRON
easat-9471	17	12	integrates	integrate	VERB
easat-9471	17	13	markov	markov	NOUN
easat-9471	17	14	chains	chain	NOUN
easat-9471	17	15	and	and	CCONJ
easat-9471	17	16	matrix	matrix	NOUN
easat-9471	17	17	factorization	factorization	NOUN
easat-9471	17	18	to	to	PART
easat-9471	17	19	capture	capture	VERB
easat-9471	17	20	long	long	ADJ
easat-9471	17	21	-	-	PUNCT
easat-9471	17	22	term	term	NOUN
easat-9471	17	23	and	and	CCONJ
easat-9471	17	24	short	short	ADJ
easat-9471	17	25	-	-	PUNCT
easat-9471	17	26	term	term	NOUN
easat-9471	17	27	preferences	preference	NOUN
easat-9471	17	28	;	;	PUNCT
easat-9471	17	29	rnn	rnn	NOUN
easat-9471	17	30	and	and	CCONJ
easat-9471	17	31	lstm	lstm	NOUN
easat-9471	17	32	-	-	PUNCT
easat-9471	17	33	based	base	VERB
easat-9471	17	34	session	session	NOUN
easat-9471	17	35	models	model	NOUN
easat-9471	17	36	;	;	PUNCT
easat-9471	17	37	gru4rec	gru4rec	PROPN
easat-9471	18	1	[	[	X
easat-9471	18	2	2	2	X
easat-9471	18	3	]	]	PUNCT
easat-9471	18	4	which	which	PRON
easat-9471	18	5	addresses	address	VERB
easat-9471	18	6	long	long	ADJ
easat-9471	18	7	-	-	PUNCT
easat-9471	18	8	term	term	NOUN
easat-9471	18	9	dependency	dependency	NOUN
easat-9471	18	10	issues	issue	NOUN
easat-9471	18	11	prior	prior	ADV
easat-9471	18	12	to	to	ADP
easat-9471	18	13	the	the	DET
easat-9471	18	14	advent	advent	NOUN
easat-9471	18	15	of	of	ADP
easat-9471	18	16	the	the	DET
easat-9471	18	17	transformer	transformer	NOUN
easat-9471	18	18	architecture	architecture	NOUN
easat-9471	18	19	;	;	PUNCT
easat-9471	18	20	and	and	CCONJ
easat-9471	18	21	sasrec	sasrec	NOUN
easat-9471	19	1	[	[	X
easat-9471	19	2	3	3	X
easat-9471	19	3	]	]	PUNCT
easat-9471	19	4	which	which	PRON
easat-9471	19	5	introduced	introduce	VERB
easat-9471	19	6	the	the	DET
easat-9471	19	7	selfattention	selfattention	NOUN
easat-9471	19	8	mechanism	mechanism	NOUN
easat-9471	19	9	from	from	ADP
easat-9471	19	10	the	the	DET
easat-9471	19	11	transformer	transformer	NOUN
easat-9471	19	12	architecture	architecture	NOUN
easat-9471	19	13	.	.	PUNCT
easat-9471	20	1	meanwhile	meanwhile	ADV
easat-9471	20	2	,	,	PUNCT
easat-9471	20	3	graph	graph	NOUN
easat-9471	20	4	-	-	PUNCT
easat-9471	20	5	based	base	VERB
easat-9471	20	6	collaborative	collaborative	ADJ
easat-9471	20	7	filtering	filtering	NOUN
easat-9471	20	8	methods	method	NOUN
easat-9471	20	9	represent	represent	VERB
easat-9471	20	10	interactions	interaction	NOUN
easat-9471	20	11	between	between	ADP
easat-9471	20	12	users	user	NOUN
easat-9471	20	13	and	and	CCONJ
easat-9471	20	14	items	item	NOUN
easat-9471	20	15	as	as	ADP
easat-9471	20	16	a	a	DET
easat-9471	20	17	bipartite	bipartite	NOUN
easat-9471	20	18	graph	graph	NOUN
easat-9471	20	19	structure	structure	NOUN
easat-9471	20	20	and	and	CCONJ
easat-9471	20	21	reinforce	reinforce	VERB
easat-9471	20	22	the	the	DET
easat-9471	20	23	sparsity	sparsity	NOUN
easat-9471	20	24	of	of	ADP
easat-9471	20	25	the	the	DET
easat-9471	20	26	bipartite	bipartite	PROPN
easat-9471	20	27	graph	graph	NOUN
easat-9471	20	28	through	through	ADP
easat-9471	20	29	multi	multi	ADJ
easat-9471	20	30	-	-	ADJ
easat-9471	20	31	layer	layer	ADJ
easat-9471	20	32	propagation	propagation	NOUN
easat-9471	20	33	.	.	PUNCT
easat-9471	21	1	for	for	ADP
easat-9471	21	2	example	example	NOUN
easat-9471	21	3	,	,	PUNCT
easat-9471	21	4	ngcf	ngcf	VERB
easat-9471	22	1	[	[	X
easat-9471	22	2	4	4	X
easat-9471	22	3	]	]	PUNCT
easat-9471	22	4	learned	learn	VERB
easat-9471	22	5	local	local	ADJ
easat-9471	22	6	neighborhood	neighborhood	NOUN
easat-9471	22	7	information	information	NOUN
easat-9471	22	8	and	and	CCONJ
easat-9471	22	9	global	global	ADJ
easat-9471	22	10	collaborative	collaborative	ADJ
easat-9471	22	11	signals	signal	NOUN
easat-9471	22	12	using	use	VERB
easat-9471	22	13	a	a	DET
easat-9471	22	14	graph	graph	NOUN
easat-9471	22	15	convolutional	convolutional	ADJ
easat-9471	22	16	network	network	NOUN
easat-9471	22	17	to	to	PART
easat-9471	22	18	capture	capture	VERB
easat-9471	22	19	higher	high	ADJ
easat-9471	22	20	-	-	PUNCT
easat-9471	22	21	order	order	NOUN
easat-9471	22	22	https://orcid.org/0009-0009-9752-4516	https://orcid.org/0009-0009-9752-4516	PROPN
easat-9471	22	23	https://orcid.org/0009-0002-6987-9634	https://orcid.org/0009-0002-6987-9634	PROPN
easat-9471	22	24	https://orcid.org/0009-0004-6277-2513	https://orcid.org/0009-0004-6277-2513	PROPN
easat-9471	22	25	https://orcid.org/0009-0003-9108-3429	https://orcid.org/0009-0003-9108-3429	VERB
easat-9471	22	26	872	872	NUM
easat-9471	22	27	edelweiss	edelweiss	PROPN
easat-9471	22	28	applied	apply	VERB
easat-9471	22	29	science	science	NOUN
easat-9471	22	30	and	and	CCONJ
easat-9471	22	31	technology	technology	NOUN
easat-9471	22	32	issn	issn	PROPN
easat-9471	22	33	:	:	PUNCT
easat-9471	22	34	2576	2576	NUM
easat-9471	22	35	-	-	SYM
easat-9471	22	36	8484	8484	NUM
easat-9471	22	37	vol	vol	NOUN
easat-9471	22	38	.	.	PROPN
easat-9471	23	1	9	9	NUM
easat-9471	23	2	,	,	PUNCT
easat-9471	23	3	no	no	INTJ
easat-9471	23	4	.	.	NOUN
easat-9471	23	5	8	8	NUM
easat-9471	23	6	:	:	PUNCT
easat-9471	23	7	871	871	NUM
easat-9471	23	8	-	-	SYM
easat-9471	23	9	881	881	NUM
easat-9471	23	10	,	,	PUNCT
easat-9471	23	11	2025	2025	NUM
easat-9471	23	12	doi	doi	NOUN
easat-9471	23	13	:	:	PUNCT
easat-9471	23	14	10.55214/2576	10.55214/2576	NUM
easat-9471	23	15	-	-	PUNCT
easat-9471	23	16	8484.v9i8.9471	8484.v9i8.9471	NOUN
easat-9471	24	1	©	©	PROPN
easat-9471	24	2	2025	2025	NUM
easat-9471	24	3	by	by	ADP
easat-9471	24	4	the	the	DET
easat-9471	24	5	authors	author	NOUN
easat-9471	24	6	;	;	PUNCT
easat-9471	24	7	licensee	licensee	PROPN
easat-9471	24	8	learning	learn	VERB
easat-9471	24	9	gate	gate	NOUN
easat-9471	24	10	connectivity	connectivity	NOUN
easat-9471	24	11	.	.	PUNCT
easat-9471	25	1	lightgcn	lightgcn	PROPN
easat-9471	26	1	[	[	X
easat-9471	26	2	5	5	NUM
easat-9471	26	3	]	]	PUNCT
easat-9471	26	4	improves	improve	VERB
easat-9471	26	5	efficiency	efficiency	NOUN
easat-9471	26	6	and	and	CCONJ
easat-9471	26	7	performance	performance	NOUN
easat-9471	26	8	by	by	ADP
easat-9471	26	9	removing	remove	VERB
easat-9471	26	10	the	the	DET
easat-9471	26	11	nonlinear	nonlinear	ADJ
easat-9471	26	12	activation	activation	NOUN
easat-9471	26	13	of	of	ADP
easat-9471	26	14	ngcf	ngcf	PROPN
easat-9471	26	15	.	.	PUNCT
easat-9471	27	1	however	however	ADV
easat-9471	27	2	,	,	PUNCT
easat-9471	27	3	when	when	SCONJ
easat-9471	27	4	sequential	sequential	ADJ
easat-9471	27	5	collaborative	collaborative	ADJ
easat-9471	27	6	filtering	filtering	NOUN
easat-9471	27	7	methods	method	NOUN
easat-9471	27	8	and	and	CCONJ
easat-9471	27	9	graph	graph	NOUN
easat-9471	27	10	-	-	PUNCT
easat-9471	27	11	based	base	VERB
easat-9471	27	12	collaborative	collaborative	ADJ
easat-9471	27	13	filtering	filtering	NOUN
easat-9471	27	14	methods	method	NOUN
easat-9471	27	15	are	be	AUX
easat-9471	27	16	used	use	VERB
easat-9471	27	17	independently	independently	ADV
easat-9471	27	18	,	,	PUNCT
easat-9471	27	19	only	only	ADV
easat-9471	27	20	their	their	PRON
easat-9471	27	21	respective	respective	ADJ
easat-9471	27	22	advantages	advantage	NOUN
easat-9471	27	23	are	be	AUX
easat-9471	27	24	utilized	utilize	VERB
easat-9471	27	25	.	.	PUNCT
easat-9471	28	1	therefore	therefore	ADV
easat-9471	28	2	,	,	PUNCT
easat-9471	28	3	recent	recent	ADJ
easat-9471	28	4	studies	study	NOUN
easat-9471	28	5	[	[	X
easat-9471	28	6	6	6	NUM
easat-9471	28	7	,	,	PUNCT
easat-9471	28	8	7	7	NUM
easat-9471	28	9	]	]	PUNCT
easat-9471	28	10	that	that	PRON
easat-9471	28	11	integrate	integrate	VERB
easat-9471	28	12	the	the	DET
easat-9471	28	13	two	two	NUM
easat-9471	28	14	methods	method	NOUN
easat-9471	28	15	have	have	AUX
easat-9471	28	16	attracted	attract	VERB
easat-9471	28	17	attention	attention	NOUN
easat-9471	28	18	.	.	PUNCT
easat-9471	29	1	in	in	ADP
easat-9471	29	2	particular	particular	ADJ
easat-9471	29	3	,	,	PUNCT
easat-9471	29	4	the	the	DET
easat-9471	29	5	latest	late	ADJ
easat-9471	29	6	model	model	NOUN
easat-9471	29	7	gsau	gsau	NOUN
easat-9471	29	8	[	[	X
easat-9471	29	9	8	8	NUM
easat-9471	29	10	]	]	PUNCT
easat-9471	29	11	aligns	align	VERB
easat-9471	29	12	and	and	CCONJ
easat-9471	29	13	homogenizes	homogenize	VERB
easat-9471	29	14	sequential	sequential	ADJ
easat-9471	29	15	and	and	CCONJ
easat-9471	29	16	graph	graph	NOUN
easat-9471	29	17	-	-	PUNCT
easat-9471	29	18	based	base	VERB
easat-9471	29	19	modules	module	NOUN
easat-9471	29	20	in	in	ADP
easat-9471	29	21	a	a	DET
easat-9471	29	22	single	single	ADJ
easat-9471	29	23	embedding	embed	VERB
easat-9471	29	24	space	space	NOUN
easat-9471	29	25	and	and	CCONJ
easat-9471	29	26	combines	combine	VERB
easat-9471	29	27	their	their	PRON
easat-9471	29	28	complementary	complementary	ADJ
easat-9471	29	29	characteristics	characteristic	NOUN
easat-9471	29	30	to	to	PART
easat-9471	29	31	contribute	contribute	VERB
easat-9471	29	32	to	to	ADP
easat-9471	29	33	improved	improve	VERB
easat-9471	29	34	recommendation	recommendation	NOUN
easat-9471	29	35	performance	performance	NOUN
easat-9471	29	36	.	.	PUNCT
easat-9471	30	1	as	as	SCONJ
easat-9471	30	2	shown	show	VERB
easat-9471	30	3	in	in	ADP
easat-9471	30	4	figure	figure	NOUN
easat-9471	30	5	1	1	NUM
easat-9471	30	6	,	,	PUNCT
easat-9471	30	7	the	the	DET
easat-9471	30	8	gsau	gsau	ADJ
easat-9471	30	9	model	model	NOUN
easat-9471	30	10	achieved	achieve	VERB
easat-9471	30	11	significant	significant	ADJ
easat-9471	30	12	improvements	improvement	NOUN
easat-9471	30	13	in	in	ADP
easat-9471	30	14	performance	performance	NOUN
easat-9471	30	15	metrics	metric	NOUN
easat-9471	30	16	compared	compare	VERB
easat-9471	30	17	to	to	ADP
easat-9471	30	18	the	the	DET
easat-9471	30	19	traditional	traditional	ADJ
easat-9471	30	20	sequential	sequential	ADJ
easat-9471	30	21	collaborative	collaborative	ADJ
easat-9471	30	22	filtering	filtering	NOUN
easat-9471	30	23	model	model	NOUN
easat-9471	30	24	sasrec	sasrec	NOUN
easat-9471	30	25	and	and	CCONJ
easat-9471	30	26	the	the	DET
easat-9471	30	27	graph	graph	NOUN
easat-9471	30	28	-	-	PUNCT
easat-9471	30	29	based	base	VERB
easat-9471	30	30	collaborative	collaborative	ADJ
easat-9471	30	31	filtering	filtering	NOUN
easat-9471	30	32	model	model	NOUN
easat-9471	30	33	lightgcn	lightgcn	PROPN
easat-9471	30	34	.	.	PUNCT
easat-9471	31	1	figure	figure	NOUN
easat-9471	31	2	1	1	NUM
easat-9471	31	3	.	.	PUNCT
easat-9471	31	4	comparison	comparison	NOUN
easat-9471	31	5	of	of	ADP
easat-9471	31	6	collaborative	collaborative	ADJ
easat-9471	31	7	filtering	filtering	NOUN
easat-9471	31	8	model	model	NOUN
easat-9471	31	9	performance	performance	NOUN
easat-9471	31	10	using	use	VERB
easat-9471	31	11	amazon	amazon	PROPN
easat-9471	31	12	industrial	industrial	PROPN
easat-9471	31	13	&	&	CCONJ
easat-9471	31	14	scientific	scientific	ADJ
easat-9471	31	15	and	and	CCONJ
easat-9471	31	16	video	video	NOUN
easat-9471	31	17	games	game	NOUN
easat-9471	31	18	data	datum	NOUN
easat-9471	31	19	in	in	ADP
easat-9471	31	20	2023	2023	NUM
easat-9471	31	21	.	.	PUNCT
easat-9471	32	1	along	along	ADP
easat-9471	32	2	with	with	ADP
easat-9471	32	3	research	research	NOUN
easat-9471	32	4	on	on	ADP
easat-9471	32	5	integrating	integrate	VERB
easat-9471	32	6	collaborative	collaborative	ADJ
easat-9471	32	7	filtering	filtering	NOUN
easat-9471	32	8	models	model	NOUN
easat-9471	32	9	,	,	PUNCT
easat-9471	32	10	research	research	NOUN
easat-9471	32	11	on	on	ADP
easat-9471	32	12	recommender	recommender	NOUN
easat-9471	32	13	systems	system	NOUN
easat-9471	32	14	utilizing	utilize	VERB
easat-9471	32	15	reinforcement	reinforcement	NOUN
easat-9471	32	16	learning	learning	NOUN
easat-9471	32	17	[	[	X
easat-9471	32	18	9	9	NUM
easat-9471	32	19	,	,	PUNCT
easat-9471	32	20	10	10	NUM
easat-9471	32	21	]	]	PUNCT
easat-9471	32	22	is	be	AUX
easat-9471	32	23	also	also	ADV
easat-9471	32	24	continuing	continue	VERB
easat-9471	32	25	.	.	PUNCT
easat-9471	33	1	this	this	PRON
easat-9471	33	2	is	be	AUX
easat-9471	33	3	because	because	SCONJ
easat-9471	33	4	existing	exist	VERB
easat-9471	33	5	collaborative	collaborative	ADJ
easat-9471	33	6	filtering	filtering	NOUN
easat-9471	33	7	-	-	PUNCT
easat-9471	33	8	based	base	VERB
easat-9471	33	9	research	research	NOUN
easat-9471	33	10	has	have	AUX
easat-9471	33	11	focused	focus	VERB
easat-9471	33	12	only	only	ADV
easat-9471	33	13	on	on	ADP
easat-9471	33	14	one	one	NUM
easat-9471	33	15	-	-	PUNCT
easat-9471	33	16	time	time	NOUN
easat-9471	33	17	and	and	CCONJ
easat-9471	33	18	immediate	immediate	ADJ
easat-9471	33	19	predictions	prediction	NOUN
easat-9471	33	20	,	,	PUNCT
easat-9471	33	21	optimizing	optimize	VERB
easat-9471	33	22	for	for	ADP
easat-9471	33	23	short	short	ADJ
easat-9471	33	24	-	-	PUNCT
easat-9471	33	25	term	term	NOUN
easat-9471	33	26	performance	performance	NOUN
easat-9471	33	27	,	,	PUNCT
easat-9471	33	28	and	and	CCONJ
easat-9471	33	29	as	as	ADP
easat-9471	33	30	a	a	DET
easat-9471	33	31	result	result	NOUN
easat-9471	33	32	,	,	PUNCT
easat-9471	33	33	only	only	ADV
easat-9471	33	34	recommending	recommend	VERB
easat-9471	33	35	items	item	NOUN
easat-9471	33	36	from	from	ADP
easat-9471	33	37	similar	similar	ADJ
easat-9471	33	38	clusters	cluster	NOUN
easat-9471	33	39	,	,	PUNCT
easat-9471	33	40	ultimately	ultimately	ADV
easat-9471	33	41	reducing	reduce	VERB
easat-9471	33	42	user	user	NOUN
easat-9471	33	43	engagement	engagement	NOUN
easat-9471	33	44	in	in	ADP
easat-9471	33	45	the	the	DET
easat-9471	33	46	long	long	ADJ
easat-9471	33	47	term	term	NOUN
easat-9471	33	48	.	.	PUNCT
easat-9471	34	1	reinforcement	reinforcement	NOUN
easat-9471	34	2	learning	learning	NOUN
easat-9471	34	3	-	-	PUNCT
easat-9471	34	4	based	base	VERB
easat-9471	34	5	approaches	approach	NOUN
easat-9471	34	6	redefine	redefine	VERB
easat-9471	34	7	the	the	DET
easat-9471	34	8	problem	problem	NOUN
easat-9471	34	9	as	as	ADP
easat-9471	34	10	optimizing	optimize	VERB
easat-9471	34	11	user	user	NOUN
easat-9471	34	12	interactions	interaction	NOUN
easat-9471	34	13	from	from	ADP
easat-9471	34	14	a	a	DET
easat-9471	34	15	long	long	ADJ
easat-9471	34	16	-	-	PUNCT
easat-9471	34	17	term	term	NOUN
easat-9471	34	18	perspective	perspective	NOUN
easat-9471	34	19	,	,	PUNCT
easat-9471	34	20	maximizing	maximize	VERB
easat-9471	34	21	the	the	DET
easat-9471	34	22	diversity	diversity	NOUN
easat-9471	34	23	of	of	ADP
easat-9471	34	24	recommendations	recommendation	NOUN
easat-9471	34	25	and	and	CCONJ
easat-9471	34	26	user	user	NOUN
easat-9471	34	27	satisfaction	satisfaction	NOUN
easat-9471	34	28	through	through	ADP
easat-9471	34	29	exploration	exploration	NOUN
easat-9471	34	30	and	and	CCONJ
easat-9471	34	31	reinforcement	reinforcement	NOUN
easat-9471	34	32	.	.	PUNCT
easat-9471	35	1	for	for	ADP
easat-9471	35	2	example	example	NOUN
easat-9471	35	3	,	,	PUNCT
easat-9471	35	4	chen	chen	PROPN
easat-9471	35	5	,	,	PUNCT
easat-9471	35	6	et	et	PROPN
easat-9471	35	7	al	al	PROPN
easat-9471	35	8	.	.	PUNCT
easat-9471	36	1	[	[	X
easat-9471	36	2	11	11	NUM
easat-9471	36	3	]	]	PUNCT
easat-9471	36	4	successfully	successfully	ADV
easat-9471	36	5	applied	apply	VERB
easat-9471	36	6	reinforcement	reinforcement	NOUN
easat-9471	36	7	learning	learn	VERB
easat-9471	36	8	agents	agent	NOUN
easat-9471	36	9	to	to	PART
easat-9471	36	10	youtube	youtube	VERB
easat-9471	36	11	's	's	PART
easat-9471	36	12	recommender	recommender	NOUN
easat-9471	36	13	system	system	NOUN
easat-9471	36	14	,	,	PUNCT
easat-9471	36	15	proving	prove	VERB
easat-9471	36	16	its	its	PRON
easat-9471	36	17	practicality	practicality	NOUN
easat-9471	36	18	.	.	PUNCT
easat-9471	37	1	however	however	ADV
easat-9471	37	2	,	,	PUNCT
easat-9471	37	3	to	to	PART
easat-9471	37	4	address	address	VERB
easat-9471	37	5	policy	policy	NOUN
easat-9471	37	6	instability	instability	NOUN
easat-9471	37	7	caused	cause	VERB
easat-9471	37	8	by	by	ADP
easat-9471	37	9	distribution	distribution	NOUN
easat-9471	37	10	variability	variability	NOUN
easat-9471	37	11	in	in	ADP
easat-9471	37	12	off	off	ADP
easat-9471	37	13	-	-	PUNCT
easat-9471	37	14	policy	policy	NOUN
easat-9471	37	15	learning	learning	NOUN
easat-9471	37	16	,	,	PUNCT
easat-9471	37	17	subsequent	subsequent	ADJ
easat-9471	37	18	research	research	NOUN
easat-9471	37	19	[	[	X
easat-9471	37	20	12	12	NUM
easat-9471	37	21	]	]	PUNCT
easat-9471	37	22	introduced	introduce	VERB
easat-9471	37	23	the	the	DET
easat-9471	37	24	actor	actor	NOUN
easat-9471	37	25	-	-	PUNCT
easat-9471	37	26	critic	critic	NOUN
easat-9471	37	27	structure	structure	NOUN
easat-9471	37	28	.	.	PUNCT
easat-9471	38	1	in	in	ADP
easat-9471	38	2	this	this	DET
easat-9471	38	3	paper	paper	NOUN
easat-9471	38	4	,	,	PUNCT
easat-9471	38	5	we	we	PRON
easat-9471	38	6	propose	propose	VERB
easat-9471	38	7	a	a	DET
easat-9471	38	8	methodology	methodology	NOUN
easat-9471	38	9	for	for	ADP
easat-9471	38	10	applying	apply	VERB
easat-9471	38	11	deepseek	deepseek	NOUN
easat-9471	38	12	's	's	PART
easat-9471	38	13	group	group	NOUN
easat-9471	38	14	relative	relative	ADJ
easat-9471	38	15	policy	policy	NOUN
easat-9471	38	16	optimization	optimization	NOUN
easat-9471	38	17	(	(	PUNCT
easat-9471	38	18	grpo	grpo	NOUN
easat-9471	38	19	)	)	PUNCT
easat-9471	39	1	[	[	X
easat-9471	39	2	13	13	NUM
easat-9471	39	3	]	]	PUNCT
easat-9471	39	4	which	which	PRON
easat-9471	39	5	was	be	AUX
easat-9471	39	6	recently	recently	ADV
easat-9471	39	7	proposed	propose	VERB
easat-9471	39	8	in	in	ADP
easat-9471	39	9	the	the	DET
easat-9471	39	10	llm	llm	PROPN
easat-9471	39	11	field	field	NOUN
easat-9471	39	12	as	as	ADP
easat-9471	39	13	a	a	DET
easat-9471	39	14	pure	pure	ADJ
easat-9471	39	15	policy	policy	NOUN
easat-9471	39	16	-	-	PUNCT
easat-9471	39	17	based	base	VERB
easat-9471	39	18	approach	approach	NOUN
easat-9471	39	19	without	without	ADP
easat-9471	39	20	a	a	DET
easat-9471	39	21	separate	separate	ADJ
easat-9471	39	22	critic	critic	NOUN
easat-9471	39	23	network	network	NOUN
easat-9471	39	24	,	,	PUNCT
easat-9471	39	25	to	to	ADP
easat-9471	39	26	recommender	recommender	NOUN
easat-9471	39	27	systems	system	NOUN
easat-9471	39	28	.	.	PUNCT
easat-9471	40	1	we	we	PRON
easat-9471	40	2	experimentally	experimentally	ADV
easat-9471	40	3	verified	verify	VERB
easat-9471	40	4	performance	performance	NOUN
easat-9471	40	5	improvements	improvement	NOUN
easat-9471	40	6	by	by	ADP
easat-9471	40	7	applying	apply	VERB
easat-9471	40	8	the	the	DET
easat-9471	40	9	grpo	grpo	NOUN
easat-9471	40	10	algorithm	algorithm	NOUN
easat-9471	40	11	to	to	ADP
easat-9471	40	12	sequential	sequential	ADJ
easat-9471	40	13	collaborative	collaborative	ADJ
easat-9471	40	14	filtering	filtering	NOUN
easat-9471	40	15	,	,	PUNCT
easat-9471	40	16	graph	graph	NOUN
easat-9471	40	17	-	-	PUNCT
easat-9471	40	18	based	base	VERB
easat-9471	40	19	collaborative	collaborative	ADJ
easat-9471	40	20	filtering	filtering	NOUN
easat-9471	40	21	,	,	PUNCT
easat-9471	40	22	and	and	CCONJ
easat-9471	40	23	integrated	integrate	VERB
easat-9471	40	24	sequential	sequential	ADJ
easat-9471	40	25	and	and	CCONJ
easat-9471	40	26	graph	graph	NOUN
easat-9471	40	27	-	-	PUNCT
easat-9471	40	28	based	base	VERB
easat-9471	40	29	collaborative	collaborative	ADJ
easat-9471	40	30	filtering	filtering	NOUN
easat-9471	40	31	models	model	NOUN
easat-9471	40	32	.	.	PUNCT
easat-9471	41	1	the	the	DET
easat-9471	41	2	main	main	ADJ
easat-9471	41	3	contributions	contribution	NOUN
easat-9471	41	4	of	of	ADP
easat-9471	41	5	this	this	DET
easat-9471	41	6	paper	paper	NOUN
easat-9471	41	7	are	be	AUX
easat-9471	41	8	as	as	SCONJ
easat-9471	41	9	follows	follow	VERB
easat-9471	41	10	:	:	PUNCT
easat-9471	41	11	•	•	ADP
easat-9471	41	12	we	we	PRON
easat-9471	41	13	are	be	AUX
easat-9471	41	14	the	the	DET
easat-9471	41	15	first	first	ADJ
easat-9471	41	16	to	to	PART
easat-9471	41	17	apply	apply	VERB
easat-9471	41	18	the	the	DET
easat-9471	41	19	deepseek	deepseek	ADJ
easat-9471	41	20	grpo	grpo	NOUN
easat-9471	41	21	algorithm	algorithm	NOUN
easat-9471	41	22	to	to	ADP
easat-9471	41	23	a	a	DET
easat-9471	41	24	collaborative	collaborative	ADJ
easat-9471	41	25	filtering	filtering	NOUN
easat-9471	41	26	-	-	PUNCT
easat-9471	41	27	based	base	VERB
easat-9471	41	28	recommender	recommender	NOUN
easat-9471	41	29	system	system	NOUN
easat-9471	41	30	,	,	PUNCT
easat-9471	41	31	demonstrating	demonstrate	VERB
easat-9471	41	32	a	a	DET
easat-9471	41	33	significant	significant	ADJ
easat-9471	41	34	improvement	improvement	NOUN
easat-9471	41	35	in	in	ADP
easat-9471	41	36	recommendation	recommendation	NOUN
easat-9471	41	37	performance	performance	NOUN
easat-9471	41	38	.	.	PUNCT
easat-9471	42	1	873	873	NUM
easat-9471	42	2	edelweiss	edelweiss	PROPN
easat-9471	42	3	applied	apply	VERB
easat-9471	42	4	science	science	NOUN
easat-9471	42	5	and	and	CCONJ
easat-9471	42	6	technology	technology	NOUN
easat-9471	42	7	issn	issn	PROPN
easat-9471	42	8	:	:	PUNCT
easat-9471	42	9	2576	2576	NUM
easat-9471	42	10	-	-	SYM
easat-9471	42	11	8484	8484	NUM
easat-9471	42	12	vol	vol	NOUN
easat-9471	42	13	.	.	PROPN
easat-9471	43	1	9	9	NUM
easat-9471	43	2	,	,	PUNCT
easat-9471	43	3	no	no	INTJ
easat-9471	43	4	.	.	NOUN
easat-9471	43	5	8	8	NUM
easat-9471	43	6	:	:	PUNCT
easat-9471	43	7	871	871	NUM
easat-9471	43	8	-	-	SYM
easat-9471	43	9	881	881	NUM
easat-9471	43	10	,	,	PUNCT
easat-9471	43	11	2025	2025	NUM
easat-9471	43	12	doi	doi	NOUN
easat-9471	43	13	:	:	PUNCT
easat-9471	43	14	10.55214/2576	10.55214/2576	NUM
easat-9471	43	15	-	-	PUNCT
easat-9471	43	16	8484.v9i8.9471	8484.v9i8.9471	NOUN
easat-9471	44	1	©	©	PROPN
easat-9471	44	2	2025	2025	NUM
easat-9471	44	3	by	by	ADP
easat-9471	44	4	the	the	DET
easat-9471	44	5	authors	author	NOUN
easat-9471	44	6	;	;	PUNCT
easat-9471	44	7	licensee	licensee	PROPN
easat-9471	44	8	learning	learning	NOUN
easat-9471	44	9	gate	gate	PROPN
easat-9471	44	10	•	•	NOUN
easat-9471	44	11	we	we	PRON
easat-9471	44	12	performed	perform	VERB
easat-9471	44	13	policy	policy	NOUN
easat-9471	44	14	optimization	optimization	NOUN
easat-9471	44	15	for	for	ADP
easat-9471	44	16	each	each	DET
easat-9471	44	17	embedding	embed	VERB
easat-9471	44	18	structure	structure	NOUN
easat-9471	44	19	of	of	ADP
easat-9471	44	20	collaborative	collaborative	ADJ
easat-9471	44	21	filtering	filtering	NOUN
easat-9471	44	22	models	model	NOUN
easat-9471	44	23	and	and	CCONJ
easat-9471	44	24	experimentally	experimentally	ADV
easat-9471	44	25	analyzed	analyze	VERB
easat-9471	44	26	and	and	CCONJ
easat-9471	44	27	verified	verify	VERB
easat-9471	44	28	the	the	DET
easat-9471	44	29	differences	difference	NOUN
easat-9471	44	30	in	in	ADP
easat-9471	44	31	performance	performance	NOUN
easat-9471	44	32	.	.	PUNCT
easat-9471	45	1	2	2	X
easat-9471	45	2	.	.	X
easat-9471	45	3	related	relate	VERB
easat-9471	45	4	work	work	NOUN
easat-9471	45	5	2.1	2.1	NUM
easat-9471	45	6	.	.	PUNCT
easat-9471	46	1	collaborative	collaborative	ADJ
easat-9471	46	2	filtering	filtering	NOUN
easat-9471	46	3	integration	integration	NOUN
easat-9471	46	4	researchers	researcher	NOUN
easat-9471	46	5	have	have	AUX
easat-9471	46	6	attempted	attempt	VERB
easat-9471	46	7	to	to	PART
easat-9471	46	8	integrate	integrate	VERB
easat-9471	46	9	sequential	sequential	ADJ
easat-9471	46	10	and	and	CCONJ
easat-9471	46	11	graph	graph	NOUN
easat-9471	46	12	-	-	PUNCT
easat-9471	46	13	based	base	VERB
easat-9471	46	14	methods	method	NOUN
easat-9471	46	15	to	to	PART
easat-9471	46	16	effectively	effectively	ADV
easat-9471	46	17	capture	capture	VERB
easat-9471	46	18	user	user	NOUN
easat-9471	46	19	’s	’s	PART
easat-9471	46	20	short	short	ADJ
easat-9471	46	21	-	-	PUNCT
easat-9471	46	22	term	term	NOUN
easat-9471	46	23	and	and	CCONJ
easat-9471	46	24	long	long	ADJ
easat-9471	46	25	-	-	PUNCT
easat-9471	46	26	term	term	NOUN
easat-9471	46	27	preferences	preference	NOUN
easat-9471	46	28	.	.	PUNCT
easat-9471	47	1	sr	sr	PROPN
easat-9471	47	2	-	-	PUNCT
easat-9471	47	3	gnn	gnn	PROPN
easat-9471	47	4	[	[	X
easat-9471	47	5	6	6	NUM
easat-9471	47	6	]	]	PUNCT
easat-9471	47	7	learns	learn	VERB
easat-9471	47	8	sequential	sequential	ADJ
easat-9471	47	9	relationships	relationship	NOUN
easat-9471	47	10	between	between	ADP
easat-9471	47	11	items	item	NOUN
easat-9471	47	12	within	within	ADP
easat-9471	47	13	a	a	DET
easat-9471	47	14	session	session	NOUN
easat-9471	47	15	using	use	VERB
easat-9471	47	16	a	a	DET
easat-9471	47	17	graph	graph	NOUN
easat-9471	47	18	neural	neural	ADJ
easat-9471	47	19	network	network	NOUN
easat-9471	47	20	and	and	CCONJ
easat-9471	47	21	predicts	predict	VERB
easat-9471	47	22	the	the	DET
easat-9471	47	23	next	next	ADJ
easat-9471	47	24	item	item	NOUN
easat-9471	47	25	by	by	ADP
easat-9471	47	26	combining	combine	VERB
easat-9471	47	27	global	global	ADJ
easat-9471	47	28	collaborative	collaborative	ADJ
easat-9471	47	29	signals	signal	NOUN
easat-9471	47	30	and	and	CCONJ
easat-9471	47	31	session	session	NOUN
easat-9471	47	32	interests	interest	NOUN
easat-9471	47	33	based	base	VERB
easat-9471	47	34	on	on	ADP
easat-9471	47	35	the	the	DET
easat-9471	47	36	last	last	ADJ
easat-9471	47	37	node	node	NOUN
easat-9471	47	38	embedding	embed	VERB
easat-9471	47	39	using	use	VERB
easat-9471	47	40	attention	attention	NOUN
easat-9471	47	41	.	.	PUNCT
easat-9471	48	1	fgnn	fgnn	NOUN
easat-9471	49	1	[	[	X
easat-9471	49	2	14	14	NUM
easat-9471	49	3	]	]	PUNCT
easat-9471	49	4	improves	improve	VERB
easat-9471	49	5	recommendation	recommendation	NOUN
easat-9471	49	6	accuracy	accuracy	NOUN
easat-9471	49	7	by	by	ADP
easat-9471	49	8	adding	add	VERB
easat-9471	49	9	a	a	DET
easat-9471	49	10	weighted	weight	VERB
easat-9471	49	11	graph	graph	NOUN
easat-9471	49	12	attention	attention	NOUN
easat-9471	49	13	layer	layer	NOUN
easat-9471	49	14	that	that	PRON
easat-9471	49	15	reflects	reflect	VERB
easat-9471	49	16	the	the	DET
easat-9471	49	17	directionality	directionality	NOUN
easat-9471	49	18	and	and	CCONJ
easat-9471	49	19	interaction	interaction	NOUN
easat-9471	49	20	frequency	frequency	NOUN
easat-9471	49	21	between	between	ADP
easat-9471	49	22	neighboring	neighboring	NOUN
easat-9471	49	23	nodes	node	NOUN
easat-9471	49	24	,	,	PUNCT
easat-9471	49	25	based	base	VERB
easat-9471	49	26	on	on	ADP
easat-9471	49	27	the	the	DET
easat-9471	49	28	structure	structure	NOUN
easat-9471	49	29	of	of	ADP
easat-9471	49	30	sr	sr	PROPN
easat-9471	49	31	-	-	PUNCT
easat-9471	49	32	gnn	gnn	PROPN
easat-9471	49	33	.	.	PUNCT
easat-9471	49	34	tagnn	tagnn	NOUN
easat-9471	50	1	[	[	X
easat-9471	50	2	15	15	NUM
easat-9471	50	3	]	]	PUNCT
easat-9471	50	4	adds	add	VERB
easat-9471	50	5	a	a	DET
easat-9471	50	6	target	target	NOUN
easat-9471	50	7	attentive	attentive	ADJ
easat-9471	50	8	module	module	NOUN
easat-9471	50	9	to	to	ADP
easat-9471	50	10	the	the	DET
easat-9471	50	11	graph	graph	NOUN
easat-9471	50	12	within	within	ADP
easat-9471	50	13	a	a	DET
easat-9471	50	14	session	session	NOUN
easat-9471	50	15	,	,	PUNCT
easat-9471	50	16	re	re	VERB
easat-9471	50	17	-	-	VERB
easat-9471	50	18	weighting	weight	VERB
easat-9471	50	19	to	to	PART
easat-9471	50	20	focus	focus	VERB
easat-9471	50	21	more	more	ADV
easat-9471	50	22	on	on	ADP
easat-9471	50	23	past	past	ADJ
easat-9471	50	24	behaviors	behavior	NOUN
easat-9471	50	25	associated	associate	VERB
easat-9471	50	26	with	with	ADP
easat-9471	50	27	the	the	DET
easat-9471	50	28	target	target	NOUN
easat-9471	50	29	item	item	NOUN
easat-9471	50	30	.	.	PUNCT
easat-9471	51	1	while	while	SCONJ
easat-9471	51	2	previous	previous	ADJ
easat-9471	51	3	studies	study	NOUN
easat-9471	51	4	attempted	attempt	VERB
easat-9471	51	5	to	to	PART
easat-9471	51	6	integrate	integrate	VERB
easat-9471	51	7	sequential	sequential	ADJ
easat-9471	51	8	and	and	CCONJ
easat-9471	51	9	graph	graph	NOUN
easat-9471	51	10	structures	structure	NOUN
easat-9471	51	11	within	within	ADP
easat-9471	51	12	a	a	DET
easat-9471	51	13	session	session	NOUN
easat-9471	51	14	,	,	PUNCT
easat-9471	51	15	cotrec	cotrec	ADV
easat-9471	51	16	[	[	X
easat-9471	51	17	16	16	NUM
easat-9471	51	18	]	]	PUNCT
easat-9471	51	19	separated	separate	VERB
easat-9471	51	20	the	the	DET
easat-9471	51	21	session	session	NOUN
easat-9471	51	22	into	into	ADP
easat-9471	51	23	internal	internal	ADJ
easat-9471	51	24	and	and	CCONJ
easat-9471	51	25	external	external	ADJ
easat-9471	51	26	graphs	graph	NOUN
easat-9471	51	27	and	and	CCONJ
easat-9471	51	28	applied	apply	VERB
easat-9471	51	29	contrastive	contrastive	ADJ
easat-9471	51	30	learning	learning	NOUN
easat-9471	51	31	between	between	ADP
easat-9471	51	32	the	the	DET
easat-9471	51	33	two	two	NUM
easat-9471	51	34	encoders	encoder	NOUN
easat-9471	51	35	to	to	PART
easat-9471	51	36	effectively	effectively	ADV
easat-9471	51	37	improve	improve	VERB
easat-9471	51	38	data	datum	NOUN
easat-9471	51	39	sparsity	sparsity	NOUN
easat-9471	51	40	issues	issue	NOUN
easat-9471	51	41	.	.	PUNCT
easat-9471	52	1	gce	gce	PROPN
easat-9471	52	2	-	-	PUNCT
easat-9471	52	3	gnn	gnn	PROPN
easat-9471	52	4	[	[	X
easat-9471	52	5	17	17	NUM
easat-9471	52	6	]	]	PUNCT
easat-9471	52	7	proposed	propose	VERB
easat-9471	52	8	a	a	DET
easat-9471	52	9	dual	dual	ADJ
easat-9471	52	10	network	network	NOUN
easat-9471	52	11	structure	structure	NOUN
easat-9471	52	12	separated	separate	VERB
easat-9471	52	13	into	into	ADP
easat-9471	52	14	session	session	NOUN
easat-9471	52	15	graphs	graph	NOUN
easat-9471	52	16	and	and	CCONJ
easat-9471	52	17	global	global	ADJ
easat-9471	52	18	graphs	graph	NOUN
easat-9471	52	19	.	.	PUNCT
easat-9471	53	1	the	the	DET
easat-9471	53	2	former	former	ADJ
easat-9471	53	3	learns	learn	VERB
easat-9471	53	4	sequential	sequential	ADJ
easat-9471	53	5	relationships	relationship	NOUN
easat-9471	53	6	between	between	ADP
easat-9471	53	7	items	item	NOUN
easat-9471	53	8	within	within	ADP
easat-9471	53	9	the	the	DET
easat-9471	53	10	current	current	ADJ
easat-9471	53	11	session	session	NOUN
easat-9471	53	12	,	,	PUNCT
easat-9471	53	13	while	while	SCONJ
easat-9471	53	14	the	the	DET
easat-9471	53	15	latter	latter	NOUN
easat-9471	53	16	learns	learn	VERB
easat-9471	53	17	sequential	sequential	ADJ
easat-9471	53	18	relationships	relationship	NOUN
easat-9471	53	19	between	between	ADP
easat-9471	53	20	items	item	NOUN
easat-9471	53	21	globally	globally	ADV
easat-9471	53	22	,	,	PUNCT
easat-9471	53	23	demonstrating	demonstrate	VERB
easat-9471	53	24	consistent	consistent	ADJ
easat-9471	53	25	performance	performance	NOUN
easat-9471	53	26	advantages	advantage	NOUN
easat-9471	53	27	across	across	ADP
easat-9471	53	28	all	all	DET
easat-9471	53	29	benchmark	benchmark	ADJ
easat-9471	53	30	datasets	dataset	NOUN
easat-9471	53	31	.	.	PUNCT
easat-9471	54	1	gag	gag	NOUN
easat-9471	55	1	[	[	X
easat-9471	55	2	18	18	NUM
easat-9471	55	3	]	]	PUNCT
easat-9471	55	4	added	add	VERB
easat-9471	55	5	user	user	NOUN
easat-9471	55	6	embeddings	embedding	NOUN
easat-9471	55	7	as	as	ADP
easat-9471	55	8	global	global	ADJ
easat-9471	55	9	features	feature	NOUN
easat-9471	55	10	to	to	ADP
easat-9471	55	11	both	both	DET
easat-9471	55	12	graphs	graph	NOUN
easat-9471	55	13	,	,	PUNCT
easat-9471	55	14	simultaneously	simultaneously	ADV
easat-9471	55	15	reflecting	reflect	VERB
easat-9471	55	16	sequential	sequential	ADJ
easat-9471	55	17	relationships	relationship	NOUN
easat-9471	55	18	between	between	ADP
easat-9471	55	19	items	item	NOUN
easat-9471	55	20	within	within	ADP
easat-9471	55	21	the	the	DET
easat-9471	55	22	current	current	ADJ
easat-9471	55	23	session	session	NOUN
easat-9471	55	24	and	and	CCONJ
easat-9471	55	25	global	global	ADJ
easat-9471	55	26	user	user	NOUN
easat-9471	55	27	preferences	preference	NOUN
easat-9471	55	28	.	.	PUNCT
easat-9471	56	1	however	however	ADV
easat-9471	56	2	,	,	PUNCT
easat-9471	56	3	sequential	sequential	ADJ
easat-9471	56	4	and	and	CCONJ
easat-9471	56	5	graph	graph	NOUN
easat-9471	56	6	integration	integration	NOUN
easat-9471	56	7	models	model	NOUN
easat-9471	56	8	are	be	AUX
easat-9471	56	9	optimized	optimize	VERB
easat-9471	56	10	for	for	ADP
easat-9471	56	11	predicting	predict	VERB
easat-9471	56	12	the	the	DET
easat-9471	56	13	next	next	ADJ
easat-9471	56	14	item	item	NOUN
easat-9471	56	15	,	,	PUNCT
easat-9471	56	16	and	and	CCONJ
easat-9471	56	17	thus	thus	ADV
easat-9471	56	18	have	have	VERB
easat-9471	56	19	limitations	limitation	NOUN
easat-9471	56	20	in	in	ADP
easat-9471	56	21	reflecting	reflect	VERB
easat-9471	56	22	multi	multi	ADJ
easat-9471	56	23	-	-	ADJ
easat-9471	56	24	stage	stage	ADJ
easat-9471	56	25	decision	decision	NOUN
easat-9471	56	26	-	-	PUNCT
easat-9471	56	27	making	make	VERB
easat-9471	56	28	processes	process	NOUN
easat-9471	56	29	such	such	ADJ
easat-9471	56	30	as	as	ADP
easat-9471	56	31	long	long	ADJ
easat-9471	56	32	-	-	PUNCT
easat-9471	56	33	term	term	NOUN
easat-9471	56	34	user	user	NOUN
easat-9471	56	35	satisfaction	satisfaction	NOUN
easat-9471	56	36	and	and	CCONJ
easat-9471	56	37	content	content	NOUN
easat-9471	56	38	diversity	diversity	NOUN
easat-9471	56	39	.	.	PUNCT
easat-9471	57	1	2.2	2.2	NUM
easat-9471	57	2	.	.	PUNCT
easat-9471	57	3	reinforcement	reinforcement	NOUN
easat-9471	57	4	learning	learning	NOUN
easat-9471	57	5	and	and	CCONJ
easat-9471	57	6	recommender	recommender	NOUN
easat-9471	57	7	systems	system	NOUN
easat-9471	57	8	2.2.1	2.2.1	NUM
easat-9471	57	9	.	.	PUNCT
easat-9471	57	10	reinforcement	reinforcement	NOUN
easat-9471	57	11	learning	learning	NOUN
easat-9471	57	12	reinforcement	reinforcement	NOUN
easat-9471	57	13	learning	learning	NOUN
easat-9471	57	14	is	be	AUX
easat-9471	57	15	a	a	DET
easat-9471	57	16	machine	machine	NOUN
easat-9471	57	17	learning	learning	NOUN
easat-9471	57	18	paradigm	paradigm	NOUN
easat-9471	57	19	in	in	ADP
easat-9471	57	20	which	which	PRON
easat-9471	57	21	an	an	DET
easat-9471	57	22	agent	agent	NOUN
easat-9471	57	23	learns	learn	VERB
easat-9471	57	24	the	the	DET
easat-9471	57	25	optimal	optimal	ADJ
easat-9471	57	26	policy	policy	NOUN
easat-9471	57	27	that	that	PRON
easat-9471	57	28	maximizes	maximize	VERB
easat-9471	57	29	the	the	DET
easat-9471	57	30	expected	expect	VERB
easat-9471	57	31	value	value	NOUN
easat-9471	57	32	of	of	ADP
easat-9471	57	33	cumulative	cumulative	ADJ
easat-9471	57	34	rewards	reward	NOUN
easat-9471	57	35	through	through	ADP
easat-9471	57	36	interaction	interaction	NOUN
easat-9471	57	37	with	with	ADP
easat-9471	57	38	the	the	DET
easat-9471	57	39	environment	environment	NOUN
easat-9471	57	40	.	.	PUNCT
easat-9471	58	1	reinforcement	reinforcement	NOUN
easat-9471	58	2	learning	learning	NOUN
easat-9471	58	3	algorithms	algorithm	NOUN
easat-9471	58	4	are	be	AUX
easat-9471	58	5	broadly	broadly	ADV
easat-9471	58	6	categorized	categorize	VERB
easat-9471	58	7	into	into	ADP
easat-9471	58	8	value	value	NOUN
easat-9471	58	9	-	-	PUNCT
easat-9471	58	10	based	base	VERB
easat-9471	58	11	methods	method	NOUN
easat-9471	58	12	and	and	CCONJ
easat-9471	58	13	policy	policy	NOUN
easat-9471	58	14	-	-	PUNCT
easat-9471	58	15	based	base	VERB
easat-9471	58	16	methods	method	NOUN
easat-9471	58	17	.	.	PUNCT
easat-9471	59	1	the	the	DET
easat-9471	59	2	former	former	ADJ
easat-9471	59	3	includes	include	VERB
easat-9471	59	4	q	q	ADV
easat-9471	59	5	-	-	PUNCT
easat-9471	59	6	learning	learn	VERB
easat-9471	59	7	[	[	X
easat-9471	59	8	19	19	NUM
easat-9471	59	9	]	]	PUNCT
easat-9471	59	10	which	which	PRON
easat-9471	59	11	approximates	approximate	VERB
easat-9471	59	12	the	the	DET
easat-9471	59	13	optimal	optimal	ADJ
easat-9471	59	14	action	action	NOUN
easat-9471	59	15	-	-	PUNCT
easat-9471	59	16	value	value	NOUN
easat-9471	59	17	function	function	NOUN
easat-9471	59	18	using	use	VERB
easat-9471	59	19	the	the	DET
easat-9471	59	20	bellman	bellman	NOUN
easat-9471	59	21	equation	equation	NOUN
easat-9471	59	22	,	,	PUNCT
easat-9471	59	23	and	and	CCONJ
easat-9471	59	24	dqn	dqn	NOUN
easat-9471	59	25	[	[	X
easat-9471	59	26	20	20	NUM
easat-9471	59	27	]	]	PUNCT
easat-9471	59	28	which	which	PRON
easat-9471	59	29	extends	extend	VERB
easat-9471	59	30	this	this	PRON
easat-9471	59	31	to	to	ADP
easat-9471	59	32	deep	deep	ADJ
easat-9471	59	33	neural	neural	ADJ
easat-9471	59	34	networks	network	NOUN
easat-9471	59	35	.	.	PUNCT
easat-9471	60	1	the	the	DET
easat-9471	60	2	latter	latter	ADJ
easat-9471	60	3	includes	include	VERB
easat-9471	60	4	the	the	DET
easat-9471	60	5	policy	policy	NOUN
easat-9471	60	6	gradient	gradient	NOUN
easat-9471	60	7	method	method	NOUN
easat-9471	60	8	proposed	propose	VERB
easat-9471	60	9	by	by	ADP
easat-9471	60	10	sutton	sutton	PROPN
easat-9471	60	11	,	,	PUNCT
easat-9471	60	12	et	et	PROPN
easat-9471	60	13	al	al	PROPN
easat-9471	60	14	.	.	PUNCT
easat-9471	61	1	[	[	X
easat-9471	61	2	21	21	NUM
easat-9471	61	3	]	]	PUNCT
easat-9471	61	4	,	,	PUNCT
easat-9471	61	5	which	which	PRON
easat-9471	61	6	directly	directly	ADV
easat-9471	61	7	parameterizes	parameterize	VERB
easat-9471	61	8	the	the	DET
easat-9471	61	9	policy	policy	NOUN
easat-9471	61	10	and	and	CCONJ
easat-9471	61	11	maximizes	maximize	VERB
easat-9471	61	12	the	the	DET
easat-9471	61	13	expected	expect	VERB
easat-9471	61	14	cumulative	cumulative	ADJ
easat-9471	61	15	reward	reward	NOUN
easat-9471	61	16	using	use	VERB
easat-9471	61	17	gradient	gradient	NOUN
easat-9471	61	18	ascent	ascent	NOUN
easat-9471	61	19	.	.	PUNCT
easat-9471	62	1	however	however	ADV
easat-9471	62	2	,	,	PUNCT
easat-9471	62	3	early	early	ADJ
easat-9471	62	4	policy	policy	NOUN
easat-9471	62	5	gradient	gradient	NOUN
easat-9471	62	6	methods	method	NOUN
easat-9471	62	7	relied	rely	VERB
easat-9471	62	8	on	on	ADP
easat-9471	62	9	monte	monte	PROPN
easat-9471	62	10	carlo	carlo	PROPN
easat-9471	62	11	estimation	estimation	PROPN
easat-9471	62	12	,	,	PUNCT
easat-9471	62	13	resulting	result	VERB
easat-9471	62	14	in	in	ADP
easat-9471	62	15	high	high	ADJ
easat-9471	62	16	variance	variance	NOUN
easat-9471	62	17	and	and	CCONJ
easat-9471	62	18	significantly	significantly	ADV
easat-9471	62	19	reduced	reduce	VERB
easat-9471	62	20	learning	learn	VERB
easat-9471	62	21	stability	stability	NOUN
easat-9471	62	22	.	.	PUNCT
easat-9471	63	1	to	to	PART
easat-9471	63	2	address	address	VERB
easat-9471	63	3	these	these	DET
easat-9471	63	4	issues	issue	NOUN
easat-9471	63	5	,	,	PUNCT
easat-9471	63	6	trpo	trpo	NOUN
easat-9471	64	1	[	[	X
easat-9471	64	2	22	22	NUM
easat-9471	64	3	]	]	PUNCT
easat-9471	64	4	limited	limit	VERB
easat-9471	64	5	the	the	DET
easat-9471	64	6	policy	policy	NOUN
easat-9471	64	7	gradient	gradient	NOUN
easat-9471	64	8	update	update	NOUN
easat-9471	64	9	rate	rate	NOUN
easat-9471	64	10	using	use	VERB
easat-9471	64	11	a	a	DET
easat-9471	64	12	kl	kl	NOUN
easat-9471	64	13	divergence	divergence	NOUN
easat-9471	64	14	constraint	constraint	NOUN
easat-9471	64	15	,	,	PUNCT
easat-9471	64	16	ensuring	ensure	VERB
easat-9471	64	17	monotonic	monotonic	ADJ
easat-9471	64	18	improvement	improvement	NOUN
easat-9471	64	19	.	.	PUNCT
easat-9471	65	1	furthermore	furthermore	ADV
easat-9471	65	2	,	,	PUNCT
easat-9471	65	3	ppo	ppo	PROPN
easat-9471	65	4	[	[	X
easat-9471	65	5	23	23	NUM
easat-9471	65	6	]	]	PUNCT
easat-9471	65	7	introduced	introduce	VERB
easat-9471	65	8	a	a	DET
easat-9471	65	9	clipped	clip	VERB
easat-9471	65	10	surrogate	surrogate	ADJ
easat-9471	65	11	objective	objective	ADJ
easat-9471	65	12	function	function	NOUN
easat-9471	65	13	to	to	PART
easat-9471	65	14	suppress	suppress	VERB
easat-9471	65	15	excessive	excessive	ADJ
easat-9471	65	16	policy	policy	NOUN
easat-9471	65	17	changes	change	NOUN
easat-9471	65	18	,	,	PUNCT
easat-9471	65	19	achieving	achieve	VERB
easat-9471	65	20	learning	learn	VERB
easat-9471	65	21	stability	stability	NOUN
easat-9471	65	22	similar	similar	ADJ
easat-9471	65	23	to	to	ADP
easat-9471	65	24	trpo	trpo	NOUN
easat-9471	65	25	while	while	SCONJ
easat-9471	65	26	improving	improve	VERB
easat-9471	65	27	computational	computational	ADJ
easat-9471	65	28	efficiency	efficiency	NOUN
easat-9471	65	29	.	.	PUNCT
easat-9471	66	1	2.2.2	2.2.2	X
easat-9471	66	2	.	.	X
easat-9471	66	3	reinforcement	reinforcement	NOUN
easat-9471	66	4	learning	learning	NOUN
easat-9471	66	5	for	for	ADP
easat-9471	66	6	llms	llm	NOUN
easat-9471	66	7	a	a	DET
easat-9471	66	8	notable	notable	ADJ
easat-9471	66	9	development	development	NOUN
easat-9471	66	10	in	in	ADP
easat-9471	66	11	recent	recent	ADJ
easat-9471	66	12	reinforcement	reinforcement	NOUN
easat-9471	66	13	learning	learning	NOUN
easat-9471	66	14	research	research	NOUN
easat-9471	66	15	is	be	AUX
easat-9471	66	16	that	that	SCONJ
easat-9471	66	17	algorithms	algorithm	NOUN
easat-9471	66	18	developed	develop	VERB
easat-9471	66	19	in	in	ADP
easat-9471	66	20	different	different	ADJ
easat-9471	66	21	fields	field	NOUN
easat-9471	66	22	are	be	AUX
easat-9471	66	23	being	be	AUX
easat-9471	66	24	successfully	successfully	ADV
easat-9471	66	25	applied	apply	VERB
easat-9471	66	26	to	to	ADP
easat-9471	66	27	other	other	ADJ
easat-9471	66	28	fields	field	NOUN
easat-9471	66	29	with	with	ADP
easat-9471	66	30	structurally	structurally	ADV
easat-9471	66	31	similar	similar	ADJ
easat-9471	66	32	characteristics	characteristic	NOUN
easat-9471	66	33	.	.	PUNCT
easat-9471	67	1	openai	openai	PROPN
easat-9471	67	2	's	's	PART
easat-9471	67	3	gpt	gpt	NOUN
easat-9471	67	4	-	-	PUNCT
easat-9471	67	5	o1	o1	NOUN
easat-9471	68	1	[	[	X
easat-9471	68	2	24	24	NUM
easat-9471	68	3	]	]	PUNCT
easat-9471	68	4	utilized	utilize	VERB
easat-9471	68	5	the	the	DET
easat-9471	68	6	ppo	ppo	PROPN
easat-9471	68	7	algorithm	algorithm	NOUN
easat-9471	68	8	to	to	PART
easat-9471	68	9	optimize	optimize	VERB
easat-9471	68	10	a	a	DET
easat-9471	68	11	human	human	ADJ
easat-9471	68	12	feedback	feedback	NOUN
easat-9471	68	13	-	-	PUNCT
easat-9471	68	14	based	base	VERB
easat-9471	68	15	reward	reward	NOUN
easat-9471	68	16	model	model	NOUN
easat-9471	68	17	,	,	PUNCT
easat-9471	68	18	thereby	thereby	ADV
easat-9471	68	19	enhancing	enhance	VERB
easat-9471	68	20	its	its	PRON
easat-9471	68	21	ability	ability	NOUN
easat-9471	68	22	to	to	PART
easat-9471	68	23	generate	generate	VERB
easat-9471	68	24	responses	response	NOUN
easat-9471	68	25	aligned	align	VERB
easat-9471	68	26	with	with	ADP
easat-9471	68	27	user	user	NOUN
easat-9471	68	28	preferences	preference	NOUN
easat-9471	68	29	.	.	PUNCT
easat-9471	69	1	a	a	DET
easat-9471	69	2	particularly	particularly	ADV
easat-9471	69	3	notable	notable	ADJ
easat-9471	69	4	development	development	NOUN
easat-9471	69	5	is	be	AUX
easat-9471	69	6	the	the	DET
easat-9471	69	7	grpo	grpo	NOUN
easat-9471	69	8	algorithm	algorithm	NOUN
easat-9471	69	9	proposed	propose	VERB
easat-9471	69	10	in	in	ADP
easat-9471	69	11	deepseek	deepseek	ADJ
easat-9471	69	12	-	-	PUNCT
easat-9471	69	13	r1	r1	NOUN
easat-9471	69	14	.	.	PUNCT
easat-9471	70	1	the	the	DET
easat-9471	70	2	grpo	grpo	NOUN
easat-9471	70	3	algorithm	algorithm	NOUN
easat-9471	70	4	demonstrated	demonstrate	VERB
easat-9471	70	5	that	that	SCONJ
easat-9471	70	6	it	it	PRON
easat-9471	70	7	can	can	AUX
easat-9471	70	8	effectively	effectively	ADV
easat-9471	70	9	improve	improve	VERB
easat-9471	70	10	policy	policy	NOUN
easat-9471	70	11	performance	performance	NOUN
easat-9471	70	12	without	without	ADP
easat-9471	70	13	a	a	DET
easat-9471	70	14	separate	separate	ADJ
easat-9471	70	15	reward	reward	NOUN
easat-9471	70	16	model	model	NOUN
easat-9471	70	17	by	by	ADP
easat-9471	70	18	utilizing	utilize	VERB
easat-9471	70	19	only	only	ADJ
easat-9471	70	20	reinforcement	reinforcement	NOUN
easat-9471	70	21	learning	learning	NOUN
easat-9471	70	22	with	with	ADP
easat-9471	70	23	verifiable	verifiable	ADJ
easat-9471	70	24	rewards	reward	NOUN
easat-9471	70	25	(	(	PUNCT
easat-9471	70	26	rlvr	rlvr	NOUN
easat-9471	70	27	)	)	PUNCT
easat-9471	70	28	.	.	PUNCT
easat-9471	71	1	yu	yu	PROPN
easat-9471	71	2	,	,	PUNCT
easat-9471	71	3	et	et	PROPN
easat-9471	71	4	al	al	PROPN
easat-9471	71	5	.	.	PUNCT
easat-9471	72	1	[	[	X
easat-9471	72	2	25	25	NUM
easat-9471	72	3	]	]	PUNCT
easat-9471	72	4	proposed	propose	VERB
easat-9471	72	5	a	a	DET
easat-9471	72	6	methodology	methodology	NOUN
easat-9471	72	7	that	that	PRON
easat-9471	72	8	utilizes	utilize	VERB
easat-9471	72	9	the	the	DET
easat-9471	72	10	degree	degree	NOUN
easat-9471	72	11	of	of	ADP
easat-9471	72	12	agreement	agreement	NOUN
easat-9471	72	13	with	with	ADP
easat-9471	72	14	expert	expert	NOUN
easat-9471	72	15	reference	reference	NOUN
easat-9471	72	16	answers	answer	VERB
easat-9471	72	17	as	as	ADP
easat-9471	72	18	a	a	DET
easat-9471	72	19	binary	binary	ADJ
easat-9471	72	20	signal	signal	NOUN
easat-9471	72	21	,	,	PUNCT
easat-9471	72	22	and	and	CCONJ
easat-9471	72	23	wu	wu	PROPN
easat-9471	72	24	,	,	PUNCT
easat-9471	72	25	et	et	PROPN
easat-9471	72	26	al	al	PROPN
easat-9471	72	27	.	.	PUNCT
easat-9471	73	1	[	[	X
easat-9471	73	2	26	26	NUM
easat-9471	73	3	]	]	PUNCT
easat-9471	73	4	achieved	achieve	VERB
easat-9471	73	5	performance	performance	NOUN
easat-9471	73	6	improvements	improvement	NOUN
easat-9471	73	7	by	by	ADP
easat-9471	73	8	setting	set	VERB
easat-9471	73	9	accuracy	accuracy	NOUN
easat-9471	73	10	and	and	CCONJ
easat-9471	73	11	visual	visual	ADJ
easat-9471	73	12	quality	quality	NOUN
easat-9471	73	13	metrics	metric	NOUN
easat-9471	73	14	as	as	ADP
easat-9471	73	15	binary	binary	ADJ
easat-9471	73	16	rewards	reward	NOUN
easat-9471	73	17	in	in	ADP
easat-9471	73	18	874	874	NUM
easat-9471	73	19	edelweiss	edelweiss	PROPN
easat-9471	73	20	applied	apply	VERB
easat-9471	73	21	science	science	NOUN
easat-9471	73	22	and	and	CCONJ
easat-9471	73	23	technology	technology	NOUN
easat-9471	73	24	issn	issn	PROPN
easat-9471	73	25	:	:	PUNCT
easat-9471	73	26	2576	2576	NUM
easat-9471	73	27	-	-	SYM
easat-9471	73	28	8484	8484	NUM
easat-9471	73	29	vol	vol	NOUN
easat-9471	73	30	.	.	PROPN
easat-9471	74	1	9	9	NUM
easat-9471	74	2	,	,	PUNCT
easat-9471	74	3	no	no	INTJ
easat-9471	74	4	.	.	NOUN
easat-9471	74	5	8	8	NUM
easat-9471	74	6	:	:	PUNCT
easat-9471	74	7	871	871	NUM
easat-9471	74	8	-	-	SYM
easat-9471	74	9	881	881	NUM
easat-9471	74	10	,	,	PUNCT
easat-9471	74	11	2025	2025	NUM
easat-9471	74	12	doi	doi	NOUN
easat-9471	74	13	:	:	PUNCT
easat-9471	74	14	10.55214/2576	10.55214/2576	NUM
easat-9471	74	15	-	-	PUNCT
easat-9471	74	16	8484.v9i8.9471	8484.v9i8.9471	NOUN
easat-9471	75	1	©	©	PROPN
easat-9471	75	2	2025	2025	NUM
easat-9471	75	3	by	by	ADP
easat-9471	75	4	the	the	DET
easat-9471	75	5	authors	author	NOUN
easat-9471	75	6	;	;	PUNCT
easat-9471	75	7	licensee	licensee	PROPN
easat-9471	75	8	learning	learning	NOUN
easat-9471	75	9	gate	gate	VERB
easat-9471	75	10	the	the	DET
easat-9471	75	11	rlvr	rlvr	NOUN
easat-9471	75	12	-	-	PUNCT
easat-9471	75	13	world	world	NOUN
easat-9471	75	14	environment	environment	NOUN
easat-9471	75	15	.	.	PUNCT
easat-9471	76	1	2.2.3	2.2.3	X
easat-9471	76	2	.	.	X
easat-9471	76	3	reinforcement	reinforcement	NOUN
easat-9471	76	4	learning	learning	NOUN
easat-9471	76	5	for	for	ADP
easat-9471	76	6	recommender	recommender	NOUN
easat-9471	76	7	systems	system	NOUN
easat-9471	76	8	reinforcement	reinforcement	NOUN
easat-9471	76	9	learning	learning	NOUN
easat-9471	76	10	in	in	ADP
easat-9471	76	11	recommender	recommender	NOUN
easat-9471	76	12	systems	system	NOUN
easat-9471	76	13	probabilistically	probabilistically	ADV
easat-9471	76	14	models	model	VERB
easat-9471	76	15	the	the	DET
easat-9471	76	16	uncertainty	uncertainty	NOUN
easat-9471	76	17	of	of	ADP
easat-9471	76	18	user	user	NOUN
easat-9471	76	19	behavior	behavior	NOUN
easat-9471	76	20	and	and	CCONJ
easat-9471	76	21	optimizes	optimize	VERB
easat-9471	76	22	long	long	ADJ
easat-9471	76	23	-	-	PUNCT
easat-9471	76	24	term	term	NOUN
easat-9471	76	25	user	user	NOUN
easat-9471	76	26	satisfaction	satisfaction	NOUN
easat-9471	76	27	and	and	CCONJ
easat-9471	76	28	engagement	engagement	NOUN
easat-9471	76	29	by	by	ADP
easat-9471	76	30	designing	design	VERB
easat-9471	76	31	multidimensional	multidimensional	ADJ
easat-9471	76	32	reward	reward	NOUN
easat-9471	76	33	functions	function	NOUN
easat-9471	76	34	such	such	ADJ
easat-9471	76	35	as	as	ADP
easat-9471	76	36	click	click	NOUN
easat-9471	76	37	-	-	PUNCT
easat-9471	76	38	through	through	ADP
easat-9471	76	39	rates	rate	NOUN
easat-9471	76	40	,	,	PUNCT
easat-9471	76	41	purchase	purchase	NOUN
easat-9471	76	42	conversion	conversion	NOUN
easat-9471	76	43	rates	rate	NOUN
easat-9471	76	44	,	,	PUNCT
easat-9471	76	45	and	and	CCONJ
easat-9471	76	46	dwell	dwell	NOUN
easat-9471	76	47	time	time	NOUN
easat-9471	76	48	.	.	PUNCT
easat-9471	77	1	this	this	DET
easat-9471	77	2	approach	approach	NOUN
easat-9471	77	3	differs	differ	VERB
easat-9471	77	4	from	from	ADP
easat-9471	77	5	traditional	traditional	ADJ
easat-9471	77	6	supervised	supervised	ADJ
easat-9471	77	7	learning	learning	NOUN
easat-9471	77	8	-	-	PUNCT
easat-9471	77	9	based	base	VERB
easat-9471	77	10	recommendation	recommendation	NOUN
easat-9471	77	11	methods	method	NOUN
easat-9471	77	12	,	,	PUNCT
easat-9471	77	13	which	which	PRON
easat-9471	77	14	focus	focus	VERB
easat-9471	77	15	on	on	ADP
easat-9471	77	16	short	short	ADJ
easat-9471	77	17	-	-	PUNCT
easat-9471	77	18	term	term	NOUN
easat-9471	77	19	prediction	prediction	NOUN
easat-9471	77	20	accuracy	accuracy	NOUN
easat-9471	77	21	.	.	PUNCT
easat-9471	78	1	prl	prl	PROPN
easat-9471	78	2	xin	xin	PROPN
easat-9471	78	3	,	,	PUNCT
easat-9471	78	4	et	et	PROPN
easat-9471	78	5	al	al	PROPN
easat-9471	78	6	.	.	PUNCT
easat-9471	79	1	[	[	X
easat-9471	79	2	27	27	NUM
easat-9471	79	3	]	]	ADJ
easat-9471	79	4	converts	convert	VERB
easat-9471	79	5	past	past	ADP
easat-9471	79	6	interaction	interaction	NOUN
easat-9471	79	7	data	datum	NOUN
easat-9471	79	8	into	into	ADP
easat-9471	79	9	state	state	NOUN
easat-9471	79	10	-	-	PUNCT
easat-9471	79	11	cumulative	cumulative	ADJ
easat-9471	79	12	reward	reward	NOUN
easat-9471	79	13	prompts	prompt	NOUN
easat-9471	79	14	and	and	CCONJ
easat-9471	79	15	replaces	replace	VERB
easat-9471	79	16	the	the	DET
easat-9471	79	17	traditional	traditional	ADJ
easat-9471	79	18	policy	policy	NOUN
easat-9471	79	19	evaluation	evaluation	NOUN
easat-9471	79	20	step	step	NOUN
easat-9471	79	21	of	of	ADP
easat-9471	79	22	value	value	NOUN
easat-9471	79	23	function	function	NOUN
easat-9471	79	24	learning	learn	VERB
easat-9471	79	25	with	with	ADP
easat-9471	79	26	cross	cross	ADJ
easat-9471	79	27	-	-	ADJ
easat-9471	79	28	entropy	entropy	ADV
easat-9471	79	29	-	-	PUNCT
easat-9471	79	30	based	base	VERB
easat-9471	79	31	supervised	supervised	ADJ
easat-9471	79	32	learning	learning	NOUN
easat-9471	79	33	,	,	PUNCT
easat-9471	79	34	thereby	thereby	ADV
easat-9471	79	35	significantly	significantly	ADV
easat-9471	79	36	improving	improve	VERB
easat-9471	79	37	recommendation	recommendation	NOUN
easat-9471	79	38	performance	performance	NOUN
easat-9471	79	39	without	without	ADP
easat-9471	79	40	the	the	DET
easat-9471	79	41	need	need	NOUN
easat-9471	79	42	for	for	ADP
easat-9471	79	43	a	a	DET
easat-9471	79	44	separate	separate	ADJ
easat-9471	79	45	critic	critic	NOUN
easat-9471	79	46	network	network	NOUN
easat-9471	79	47	.	.	PUNCT
easat-9471	80	1	csa	csa	PROPN
easat-9471	81	1	[	[	X
easat-9471	81	2	28	28	NUM
easat-9471	81	3	]	]	PUNCT
easat-9471	81	4	improves	improve	VERB
easat-9471	81	5	the	the	DET
easat-9471	81	6	stability	stability	NOUN
easat-9471	81	7	of	of	ADP
easat-9471	81	8	policy	policy	NOUN
easat-9471	81	9	learning	learning	NOUN
easat-9471	81	10	by	by	ADP
easat-9471	81	11	integrating	integrate	VERB
easat-9471	81	12	single	single	ADJ
easat-9471	81	13	-	-	PUNCT
easat-9471	81	14	state	state	NOUN
easat-9471	81	15	data	datum	NOUN
easat-9471	81	16	augmentation	augmentation	NOUN
easat-9471	81	17	and	and	CCONJ
easat-9471	81	18	contrastive	contrastive	ADJ
easat-9471	81	19	learning	learning	NOUN
easat-9471	81	20	in	in	ADP
easat-9471	81	21	sparse	sparse	ADJ
easat-9471	81	22	reward	reward	NOUN
easat-9471	81	23	environments	environment	NOUN
easat-9471	81	24	.	.	PUNCT
easat-9471	82	1	similarly	similarly	ADV
easat-9471	82	2	,	,	PUNCT
easat-9471	82	3	mecrl	mecrl	PROPN
easat-9471	83	1	[	[	X
easat-9471	83	2	29	29	NUM
easat-9471	83	3	]	]	PUNCT
easat-9471	83	4	addresses	address	VERB
easat-9471	83	5	the	the	DET
easat-9471	83	6	data	datum	NOUN
easat-9471	83	7	scarcity	scarcity	NOUN
easat-9471	83	8	issue	issue	NOUN
easat-9471	83	9	in	in	ADP
easat-9471	83	10	sequential	sequential	ADJ
easat-9471	83	11	recommendation	recommendation	NOUN
easat-9471	83	12	environments	environment	NOUN
easat-9471	83	13	by	by	ADP
easat-9471	83	14	generating	generate	VERB
easat-9471	83	15	virtual	virtual	ADJ
easat-9471	83	16	state	state	NOUN
easat-9471	83	17	transitions	transition	NOUN
easat-9471	83	18	using	use	VERB
easat-9471	83	19	an	an	DET
easat-9471	83	20	environment	environment	NOUN
easat-9471	83	21	model	model	NOUN
easat-9471	83	22	and	and	CCONJ
easat-9471	83	23	significantly	significantly	ADV
easat-9471	83	24	improving	improve	VERB
easat-9471	83	25	the	the	DET
easat-9471	83	26	stability	stability	NOUN
easat-9471	83	27	of	of	ADP
easat-9471	83	28	policy	policy	NOUN
easat-9471	83	29	learning	learning	NOUN
easat-9471	83	30	and	and	CCONJ
easat-9471	83	31	recommendation	recommendation	NOUN
easat-9471	83	32	accuracy	accuracy	NOUN
easat-9471	83	33	through	through	ADP
easat-9471	83	34	contrastive	contrastive	ADJ
easat-9471	83	35	learning	learning	NOUN
easat-9471	83	36	.	.	PUNCT
easat-9471	84	1	however	however	ADV
easat-9471	84	2	,	,	PUNCT
easat-9471	84	3	reinforcement	reinforcement	NOUN
easat-9471	84	4	learning	learning	NOUN
easat-9471	84	5	research	research	NOUN
easat-9471	84	6	in	in	ADP
easat-9471	84	7	the	the	DET
easat-9471	84	8	existing	exist	VERB
easat-9471	84	9	recommender	recommender	NOUN
easat-9471	84	10	system	system	NOUN
easat-9471	84	11	field	field	NOUN
easat-9471	84	12	has	have	AUX
easat-9471	84	13	mainly	mainly	ADV
easat-9471	84	14	developed	develop	VERB
easat-9471	84	15	using	use	VERB
easat-9471	84	16	value	value	NOUN
easat-9471	84	17	-	-	PUNCT
easat-9471	84	18	based	base	VERB
easat-9471	84	19	methods	method	NOUN
easat-9471	84	20	or	or	CCONJ
easat-9471	84	21	limited	limited	ADJ
easat-9471	84	22	forms	form	NOUN
easat-9471	84	23	of	of	ADP
easat-9471	84	24	policy	policy	NOUN
easat-9471	84	25	gradient	gradient	NOUN
easat-9471	84	26	methods	method	NOUN
easat-9471	84	27	,	,	PUNCT
easat-9471	84	28	and	and	CCONJ
easat-9471	84	29	the	the	DET
easat-9471	84	30	application	application	NOUN
easat-9471	84	31	of	of	ADP
easat-9471	84	32	next	next	ADJ
easat-9471	84	33	-	-	PUNCT
easat-9471	84	34	generation	generation	NOUN
easat-9471	84	35	algorithms	algorithm	NOUN
easat-9471	84	36	such	such	ADJ
easat-9471	84	37	as	as	ADP
easat-9471	84	38	grpo	grpo	NOUN
easat-9471	84	39	,	,	PUNCT
easat-9471	84	40	which	which	PRON
easat-9471	84	41	has	have	AUX
easat-9471	84	42	recently	recently	ADV
easat-9471	84	43	gained	gain	VERB
easat-9471	84	44	attention	attention	NOUN
easat-9471	84	45	in	in	ADP
easat-9471	84	46	llm	llm	PROPN
easat-9471	84	47	research	research	NOUN
easat-9471	84	48	,	,	PUNCT
easat-9471	84	49	has	have	AUX
easat-9471	84	50	not	not	PART
easat-9471	84	51	yet	yet	ADV
easat-9471	84	52	been	be	AUX
easat-9471	84	53	sufficiently	sufficiently	ADV
easat-9471	84	54	studied	study	VERB
easat-9471	84	55	.	.	PUNCT
easat-9471	85	1	in	in	ADP
easat-9471	85	2	particular	particular	ADJ
easat-9471	85	3	,	,	PUNCT
easat-9471	85	4	considering	consider	VERB
easat-9471	85	5	the	the	DET
easat-9471	85	6	excellent	excellent	ADJ
easat-9471	85	7	performance	performance	NOUN
easat-9471	85	8	of	of	ADP
easat-9471	85	9	grpo	grpo	NOUN
easat-9471	85	10	in	in	ADP
easat-9471	85	11	rlvr	rlvr	ADJ
easat-9471	85	12	environments	environment	NOUN
easat-9471	85	13	proven	prove	VERB
easat-9471	85	14	in	in	ADP
easat-9471	85	15	llm	llm	PROPN
easat-9471	85	16	research	research	NOUN
easat-9471	85	17	,	,	PUNCT
easat-9471	85	18	it	it	PRON
easat-9471	85	19	is	be	AUX
easat-9471	85	20	necessary	necessary	ADJ
easat-9471	85	21	to	to	PART
easat-9471	85	22	verify	verify	VERB
easat-9471	85	23	the	the	DET
easat-9471	85	24	applicability	applicability	NOUN
easat-9471	85	25	and	and	CCONJ
easat-9471	85	26	effectiveness	effectiveness	NOUN
easat-9471	85	27	of	of	ADP
easat-9471	85	28	grpo	grpo	NOUN
easat-9471	85	29	in	in	ADP
easat-9471	85	30	the	the	DET
easat-9471	85	31	recommender	recommender	NOUN
easat-9471	85	32	systems	system	NOUN
easat-9471	85	33	field	field	NOUN
easat-9471	85	34	,	,	PUNCT
easat-9471	85	35	which	which	PRON
easat-9471	85	36	has	have	VERB
easat-9471	85	37	similar	similar	ADJ
easat-9471	85	38	structural	structural	ADJ
easat-9471	85	39	characteristics	characteristic	NOUN
easat-9471	85	40	.	.	PUNCT
easat-9471	86	1	3	3	X
easat-9471	86	2	.	.	X
easat-9471	86	3	methodology	methodology	NOUN
easat-9471	86	4	in	in	ADP
easat-9471	86	5	this	this	DET
easat-9471	86	6	section	section	NOUN
easat-9471	86	7	,	,	PUNCT
easat-9471	86	8	we	we	PRON
easat-9471	86	9	describe	describe	VERB
easat-9471	86	10	how	how	SCONJ
easat-9471	86	11	to	to	PART
easat-9471	86	12	perform	perform	VERB
easat-9471	86	13	policy	policy	NOUN
easat-9471	86	14	gradient	gradient	NOUN
easat-9471	86	15	updates	update	NOUN
easat-9471	86	16	by	by	ADP
easat-9471	86	17	applying	apply	VERB
easat-9471	86	18	the	the	DET
easat-9471	86	19	grpo	grpo	NOUN
easat-9471	86	20	algorithm	algorithm	NOUN
easat-9471	86	21	to	to	ADP
easat-9471	86	22	existing	exist	VERB
easat-9471	86	23	collaborative	collaborative	ADJ
easat-9471	86	24	filtering	filtering	NOUN
easat-9471	86	25	models	model	NOUN
easat-9471	86	26	.	.	PUNCT
easat-9471	87	1	the	the	DET
easat-9471	87	2	rationale	rationale	NOUN
easat-9471	87	3	for	for	ADP
easat-9471	87	4	applying	apply	VERB
easat-9471	87	5	the	the	DET
easat-9471	87	6	grpo	grpo	NOUN
easat-9471	87	7	algorithm	algorithm	NOUN
easat-9471	87	8	is	be	AUX
easat-9471	87	9	as	as	SCONJ
easat-9471	87	10	follows	follow	VERB
easat-9471	87	11	:	:	PUNCT
easat-9471	87	12	•	•	NUM
easat-9471	87	13	reinforcement	reinforcement	NOUN
easat-9471	87	14	learning	learning	NOUN
easat-9471	87	15	with	with	ADP
easat-9471	87	16	verifiable	verifiable	ADJ
easat-9471	87	17	rewards	reward	NOUN
easat-9471	87	18	.	.	PUNCT
easat-9471	88	1	the	the	DET
easat-9471	88	2	grpo	grpo	NOUN
easat-9471	88	3	algorithm	algorithm	PROPN
easat-9471	88	4	uses	use	VERB
easat-9471	88	5	a	a	DET
easat-9471	88	6	binary	binary	ADJ
easat-9471	88	7	reward	reward	NOUN
easat-9471	88	8	structure	structure	NOUN
easat-9471	88	9	of	of	ADP
easat-9471	88	10	acceptance	acceptance	NOUN
easat-9471	88	11	or	or	CCONJ
easat-9471	88	12	rejection	rejection	NOUN
easat-9471	88	13	for	for	ADP
easat-9471	88	14	each	each	PRON
easat-9471	88	15	token	token	VERB
easat-9471	88	16	during	during	ADP
easat-9471	88	17	the	the	DET
easat-9471	88	18	llm	llm	PROPN
easat-9471	88	19	learning	learning	NOUN
easat-9471	88	20	process	process	NOUN
easat-9471	88	21	.	.	PUNCT
easat-9471	89	1	this	this	PRON
easat-9471	89	2	can	can	AUX
easat-9471	89	3	be	be	AUX
easat-9471	89	4	interpreted	interpret	VERB
easat-9471	89	5	similarly	similarly	ADV
easat-9471	89	6	to	to	ADP
easat-9471	89	7	user	user	NOUN
easat-9471	89	8	interactions	interaction	NOUN
easat-9471	89	9	such	such	ADJ
easat-9471	89	10	as	as	ADP
easat-9471	89	11	clicks	click	NOUN
easat-9471	89	12	or	or	CCONJ
easat-9471	89	13	non	non	NOUN
easat-9471	89	14	-	-	NOUN
easat-9471	89	15	clicks	click	NOUN
easat-9471	89	16	in	in	ADP
easat-9471	89	17	recommender	recommender	NOUN
easat-9471	89	18	systems	system	NOUN
easat-9471	89	19	.	.	PUNCT
easat-9471	90	1	•	•	NUM
easat-9471	90	2	high	high	ADV
easat-9471	90	3	-	-	PUNCT
easat-9471	90	4	dimensional	dimensional	ADJ
easat-9471	90	5	action	action	NOUN
easat-9471	90	6	space	space	NOUN
easat-9471	90	7	.	.	PUNCT
easat-9471	91	1	deepseek	deepseek	ADJ
easat-9471	91	2	-	-	PUNCT
easat-9471	91	3	r1	r1	NOUN
easat-9471	91	4	selects	select	NOUN
easat-9471	91	5	and	and	CCONJ
easat-9471	91	6	generates	generate	VERB
easat-9471	91	7	tokens	token	NOUN
easat-9471	91	8	from	from	ADP
easat-9471	91	9	a	a	DET
easat-9471	91	10	vast	vast	ADJ
easat-9471	91	11	vocabulary	vocabulary	NOUN
easat-9471	91	12	.	.	PUNCT
easat-9471	92	1	this	this	PRON
easat-9471	92	2	can	can	AUX
easat-9471	92	3	be	be	AUX
easat-9471	92	4	interpreted	interpret	VERB
easat-9471	92	5	as	as	ADP
easat-9471	92	6	similar	similar	ADJ
easat-9471	92	7	to	to	ADP
easat-9471	92	8	the	the	DET
easat-9471	92	9	structure	structure	NOUN
easat-9471	92	10	of	of	ADP
easat-9471	92	11	selecting	select	VERB
easat-9471	92	12	one	one	NUM
easat-9471	92	13	recommendation	recommendation	NOUN
easat-9471	92	14	from	from	ADP
easat-9471	92	15	a	a	DET
easat-9471	92	16	vast	vast	ADJ
easat-9471	92	17	pool	pool	NOUN
easat-9471	92	18	of	of	ADP
easat-9471	92	19	recommendations	recommendation	NOUN
easat-9471	92	20	in	in	ADP
easat-9471	92	21	a	a	DET
easat-9471	92	22	recommender	recommender	NOUN
easat-9471	92	23	system	system	NOUN
easat-9471	92	24	.	.	PUNCT
easat-9471	93	1	based	base	VERB
easat-9471	93	2	on	on	ADP
easat-9471	93	3	these	these	DET
easat-9471	93	4	commonalities	commonality	NOUN
easat-9471	93	5	,	,	PUNCT
easat-9471	93	6	this	this	DET
easat-9471	93	7	section	section	NOUN
easat-9471	93	8	proposes	propose	VERB
easat-9471	93	9	a	a	DET
easat-9471	93	10	method	method	NOUN
easat-9471	93	11	for	for	ADP
easat-9471	93	12	implementing	implement	VERB
easat-9471	93	13	an	an	DET
easat-9471	93	14	effective	effective	ADJ
easat-9471	93	15	recommender	recommender	NOUN
easat-9471	93	16	system	system	NOUN
easat-9471	93	17	by	by	ADP
easat-9471	93	18	interpreting	interpret	VERB
easat-9471	93	19	user	user	NOUN
easat-9471	93	20	interactions	interaction	NOUN
easat-9471	93	21	in	in	ADP
easat-9471	93	22	collaborative	collaborative	ADJ
easat-9471	93	23	filtering	filtering	NOUN
easat-9471	93	24	as	as	ADP
easat-9471	93	25	verifiable	verifiable	ADJ
easat-9471	93	26	binary	binary	ADJ
easat-9471	93	27	reward	reward	NOUN
easat-9471	93	28	signals	signal	NOUN
easat-9471	93	29	and	and	CCONJ
easat-9471	93	30	utilizing	utilize	VERB
easat-9471	93	31	the	the	DET
easat-9471	93	32	grpo	grpo	NOUN
easat-9471	93	33	algorithm	algorithm	NOUN
easat-9471	93	34	to	to	PART
easat-9471	93	35	learn	learn	VERB
easat-9471	93	36	the	the	DET
easat-9471	93	37	optimal	optimal	ADJ
easat-9471	93	38	policy	policy	NOUN
easat-9471	93	39	from	from	ADP
easat-9471	93	40	a	a	DET
easat-9471	93	41	vast	vast	ADJ
easat-9471	93	42	pool	pool	NOUN
easat-9471	93	43	of	of	ADP
easat-9471	93	44	recommendation	recommendation	NOUN
easat-9471	93	45	candidates	candidate	NOUN
easat-9471	93	46	.	.	PUNCT
easat-9471	94	1	3.1	3.1	NUM
easat-9471	94	2	.	.	PUNCT
easat-9471	94	3	collaborative	collaborative	ADJ
easat-9471	94	4	filtering	filtering	NOUN
easat-9471	94	5	model	model	NOUN
easat-9471	94	6	3.1.1	3.1.1	NUM
easat-9471	94	7	.	.	PUNCT
easat-9471	95	1	sequential	sequential	ADJ
easat-9471	95	2	collaborative	collaborative	ADJ
easat-9471	95	3	filtering	filtering	NOUN
easat-9471	95	4	model	model	NOUN
easat-9471	95	5	sequential	sequential	PROPN
easat-9471	95	6	collaborative	collaborative	ADJ
easat-9471	95	7	filtering	filtering	NOUN
easat-9471	95	8	models	model	NOUN
easat-9471	95	9	analyze	analyze	VERB
easat-9471	95	10	historical	historical	ADJ
easat-9471	95	11	user	user	NOUN
easat-9471	95	12	behavior	behavior	NOUN
easat-9471	95	13	in	in	ADP
easat-9471	95	14	chronological	chronological	ADJ
easat-9471	95	15	order	order	NOUN
easat-9471	95	16	.	.	PUNCT
easat-9471	96	1	let	let	VERB
easat-9471	96	2	𝑢	𝑢	PRON
easat-9471	96	3	∈	∈	PROPN
easat-9471	96	4	𝒰	𝒰	PROPN
easat-9471	96	5	be	be	AUX
easat-9471	96	6	a	a	DET
easat-9471	96	7	user	user	NOUN
easat-9471	96	8	who	who	PRON
easat-9471	96	9	interacts	interact	VERB
easat-9471	96	10	with	with	ADP
easat-9471	96	11	item	item	NOUN
easat-9471	96	12	𝑖𝑡	𝑖𝑡	NOUN
easat-9471	96	13	𝑢	𝑢	X
easat-9471	96	14	∈	∈	PROPN
easat-9471	97	1	ℐ	ℐ	PRON
easat-9471	97	2	at	at	ADP
easat-9471	97	3	time	time	NOUN
easat-9471	97	4	𝑡.	𝑡.	VERB
easat-9471	97	5	the	the	DET
easat-9471	97	6	chronologically	chronologically	ADV
easat-9471	97	7	ordered	order	VERB
easat-9471	97	8	sequence	sequence	NOUN
easat-9471	97	9	of	of	ADP
easat-9471	97	10	items	item	NOUN
easat-9471	97	11	with	with	ADP
easat-9471	97	12	which	which	PRON
easat-9471	97	13	user	user	NOUN
easat-9471	97	14	𝑢	𝑢	PROPN
easat-9471	97	15	has	have	AUX
easat-9471	97	16	interacted	interact	VERB
easat-9471	97	17	is	be	AUX
easat-9471	97	18	denoted	denote	VERB
easat-9471	97	19	by𝒮𝓊	by𝒮𝓊	ADJ
easat-9471	97	20	=	=	SYM
easat-9471	97	21	(	(	PUNCT
easat-9471	97	22	𝑖1	𝑖1	PROPN
easat-9471	97	23	𝑢	𝑢	PROPN
easat-9471	97	24	,	,	PUNCT
easat-9471	97	25	  	  	SPACE
easat-9471	97	26	𝑖2	𝑖2	PROPN
easat-9471	97	27	𝑢	𝑢	PROPN
easat-9471	97	28	,	,	PUNCT
easat-9471	97	29	 	 	SPACE
easat-9471	97	30	…	…	PUNCT
easat-9471	97	31	,	,	PUNCT
easat-9471	97	32	 	 	SPACE
easat-9471	97	33	𝑖|𝒮𝓊|	𝑖|𝒮𝓊|	VERB
easat-9471	97	34	𝑢	𝑢	NOUN
easat-9471	97	35	)	)	PUNCT
easat-9471	97	36	.	.	PUNCT
easat-9471	98	1	for	for	ADP
easat-9471	98	2	every	every	DET
easat-9471	98	3	item	item	NOUN
easat-9471	98	4	𝑖	𝑖	SYM
easat-9471	98	5	∈	∈	PROPN
easat-9471	98	6	ℐ	ℐ	PROPN
easat-9471	98	7	,	,	PUNCT
easat-9471	98	8	we	we	PRON
easat-9471	98	9	define	define	VERB
easat-9471	98	10	a	a	DET
easat-9471	98	11	trainable	trainable	ADJ
easat-9471	98	12	item	item	NOUN
easat-9471	98	13	embedding	embed	VERB
easat-9471	98	14	matrix	matrix	NOUN
easat-9471	98	15	𝑀	𝑀	PROPN
easat-9471	98	16	∈	∈	PROPN
easat-9471	98	17	𝑅|𝕀|×𝑑	𝑅|𝕀|×𝑑	ADP
easat-9471	98	18	whose	whose	DET
easat-9471	98	19	rows	row	NOUN
easat-9471	98	20	are	be	AUX
easat-9471	98	21	𝑑-dimensional	𝑑-dimensional	ADJ
easat-9471	98	22	vectors	vector	NOUN
easat-9471	98	23	.	.	PUNCT
easat-9471	99	1	we	we	PRON
easat-9471	99	2	further	far	ADV
easat-9471	99	3	introduce	introduce	VERB
easat-9471	99	4	a	a	DET
easat-9471	99	5	position	position	NOUN
easat-9471	99	6	embedding	embed	VERB
easat-9471	99	7	matrix	matrix	NOUN
easat-9471	99	8	𝑃	𝑃	NOUN
easat-9471	99	9	∈	∈	NOUN
easat-9471	99	10	𝑅|𝕊	𝑅|𝕊	NOUN
easat-9471	99	11	𝕦|×𝑑	𝕦|×𝑑	NOUN
easat-9471	99	12	that	that	PRON
easat-9471	99	13	encodes	encode	VERB
easat-9471	99	14	the	the	DET
easat-9471	99	15	position	position	NOUN
easat-9471	99	16	of	of	ADP
easat-9471	99	17	each	each	DET
easat-9471	99	18	item	item	NOUN
easat-9471	99	19	in	in	ADP
easat-9471	99	20	the	the	DET
easat-9471	99	21	sequence	sequence	NOUN
easat-9471	99	22	.	.	PUNCT
easat-9471	100	1	by	by	ADP
easat-9471	100	2	selecting	select	VERB
easat-9471	100	3	the	the	DET
easat-9471	100	4	rows	row	NOUN
easat-9471	100	5	of	of	ADP
easat-9471	100	6	𝑀	𝑀	PROPN
easat-9471	100	7	that	that	PRON
easat-9471	100	8	correspond	correspond	VERB
easat-9471	100	9	to	to	ADP
easat-9471	100	10	𝒮𝓊	𝒮𝓊	PROPN
easat-9471	100	11	and	and	CCONJ
easat-9471	100	12	adding	add	VERB
easat-9471	100	13	the	the	DET
easat-9471	100	14	associated	associated	ADJ
easat-9471	100	15	position	position	NOUN
easat-9471	100	16	embeddings	embedding	NOUN
easat-9471	100	17	,	,	PUNCT
easat-9471	100	18	we	we	PRON
easat-9471	100	19	obtain	obtain	VERB
easat-9471	100	20	the	the	DET
easat-9471	100	21	sequential	sequential	ADJ
easat-9471	100	22	embeddings	embedding	NOUN
easat-9471	100	23	𝐸𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	𝐸𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	PROPN
easat-9471	100	24	as	as	SCONJ
easat-9471	100	25	follows	follow	VERB
easat-9471	100	26	:	:	PUNCT
easat-9471	100	27	875	875	NUM
easat-9471	100	28	edelweiss	edelweiss	PROPN
easat-9471	100	29	applied	apply	VERB
easat-9471	100	30	science	science	NOUN
easat-9471	100	31	and	and	CCONJ
easat-9471	100	32	technology	technology	NOUN
easat-9471	100	33	issn	issn	PROPN
easat-9471	100	34	:	:	PUNCT
easat-9471	100	35	2576	2576	NUM
easat-9471	100	36	-	-	SYM
easat-9471	100	37	8484	8484	NUM
easat-9471	100	38	vol	vol	NOUN
easat-9471	100	39	.	.	PROPN
easat-9471	101	1	9	9	NUM
easat-9471	101	2	,	,	PUNCT
easat-9471	101	3	no	no	INTJ
easat-9471	101	4	.	.	NOUN
easat-9471	101	5	8	8	NUM
easat-9471	101	6	:	:	PUNCT
easat-9471	101	7	871	871	NUM
easat-9471	101	8	-	-	SYM
easat-9471	101	9	881	881	NUM
easat-9471	101	10	,	,	PUNCT
easat-9471	101	11	2025	2025	NUM
easat-9471	101	12	doi	doi	NOUN
easat-9471	101	13	:	:	PUNCT
easat-9471	101	14	10.55214/2576	10.55214/2576	NUM
easat-9471	101	15	-	-	PUNCT
easat-9471	101	16	8484.v9i8.9471	8484.v9i8.9471	NOUN
easat-9471	102	1	©	©	PROPN
easat-9471	102	2	2025	2025	NUM
easat-9471	102	3	by	by	ADP
easat-9471	102	4	the	the	DET
easat-9471	102	5	authors	author	NOUN
easat-9471	102	6	;	;	PUNCT
easat-9471	102	7	licensee	licensee	PROPN
easat-9471	102	8	learning	learning	NOUN
easat-9471	102	9	gate	gate	NOUN
easat-9471	102	10	𝐸𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	𝐸𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	PROPN
easat-9471	102	11	=	=	PUNCT
easat-9471	102	12	[	[	PUNCT
easat-9471	102	13	𝑀𝑖1	𝑀𝑖1	NOUN
easat-9471	102	14	𝑢	𝑢	PROPN
easat-9471	102	15	+	+	PUNCT
easat-9471	102	16	𝑃1	𝑃1	NOUN
easat-9471	102	17	𝑀𝑖2	𝑀𝑖2	VERB
easat-9471	102	18	𝑢	𝑢	ADP
easat-9471	102	19	+	+	NUM
easat-9471	102	20	𝑃2	𝑃2	ADJ
easat-9471	102	21	⋮	⋮	NOUN
easat-9471	103	1	𝑀𝑖	𝑀𝑖	PROPN
easat-9471	103	2	|𝑆𝑢|−1	|𝑆𝑢|−1	NOUN
easat-9471	103	3	𝑢	𝑢	PART
easat-9471	103	4	+	+	NUM
easat-9471	103	5	𝑃|𝑆𝑢|−1𝑀𝑖	𝑃|𝑆𝑢|−1𝑀𝑖	PROPN
easat-9471	103	6	|𝑆𝑢|	|𝑆𝑢|	PROPN
easat-9471	103	7	𝑢	𝑢	PROPN
easat-9471	103	8	+	+	X
easat-9471	103	9	𝑃|𝑆𝑢|	𝑃|𝑆𝑢|	SYM
easat-9471	103	10	]	]	X
easat-9471	103	11	∈	∈	PROPN
easat-9471	103	12	𝑅|𝕊	𝑅|𝕊	NOUN
easat-9471	103	13	𝕦|×𝑑	𝕦|×𝑑	NOUN
easat-9471	103	14	(	(	PUNCT
easat-9471	103	15	1	1	X
easat-9471	103	16	)	)	PUNCT
easat-9471	103	17	the	the	DET
easat-9471	103	18	models	model	NOUN
easat-9471	103	19	that	that	PRON
easat-9471	103	20	generate	generate	VERB
easat-9471	103	21	embeddings	embedding	NOUN
easat-9471	103	22	using	use	VERB
easat-9471	103	23	the	the	DET
easat-9471	103	24	above	above	ADJ
easat-9471	103	25	architecture	architecture	NOUN
easat-9471	103	26	are	be	AUX
easat-9471	103	27	gru4rec	gru4rec	NOUN
easat-9471	103	28	,	,	PUNCT
easat-9471	103	29	which	which	PRON
easat-9471	103	30	uses	use	VERB
easat-9471	103	31	gated	gate	VERB
easat-9471	103	32	recurrent	recurrent	ADJ
easat-9471	103	33	unit	unit	NOUN
easat-9471	103	34	,	,	PUNCT
easat-9471	103	35	and	and	CCONJ
easat-9471	103	36	sasrec	sasrec	NOUN
easat-9471	103	37	,	,	PUNCT
easat-9471	103	38	which	which	PRON
easat-9471	103	39	replaces	replace	VERB
easat-9471	103	40	recurrent	recurrent	ADJ
easat-9471	103	41	layers	layer	NOUN
easat-9471	103	42	with	with	ADP
easat-9471	103	43	self	self	NOUN
easat-9471	103	44	-	-	PUNCT
easat-9471	103	45	attention	attention	NOUN
easat-9471	103	46	.	.	PUNCT
easat-9471	104	1	3.1.2	3.1.2	X
easat-9471	104	2	.	.	PUNCT
easat-9471	104	3	graph	graph	NOUN
easat-9471	104	4	-	-	PUNCT
easat-9471	104	5	based	base	VERB
easat-9471	104	6	collaborative	collaborative	ADJ
easat-9471	104	7	filtering	filtering	NOUN
easat-9471	104	8	model	model	NOUN
easat-9471	104	9	graph	graph	NOUN
easat-9471	104	10	-	-	PUNCT
easat-9471	104	11	based	base	VERB
easat-9471	104	12	collaborative	collaborative	ADJ
easat-9471	104	13	filtering	filtering	NOUN
easat-9471	104	14	models	model	NOUN
easat-9471	104	15	analyze	analyze	VERB
easat-9471	104	16	the	the	DET
easat-9471	104	17	structural	structural	ADJ
easat-9471	104	18	relationships	relationship	NOUN
easat-9471	104	19	between	between	ADP
easat-9471	104	20	users	user	NOUN
easat-9471	104	21	and	and	CCONJ
easat-9471	104	22	items	item	NOUN
easat-9471	104	23	via	via	ADP
easat-9471	104	24	graph	graph	NOUN
easat-9471	104	25	propagation	propagation	NOUN
easat-9471	104	26	.	.	PUNCT
easat-9471	105	1	given	give	VERB
easat-9471	105	2	a	a	DET
easat-9471	105	3	bipartite	bipartite	NOUN
easat-9471	105	4	graph	graph	NOUN
easat-9471	105	5	𝒢	𝒢	NOUN
easat-9471	105	6	=	=	PUNCT
easat-9471	105	7	(	(	PUNCT
easat-9471	105	8	𝒰	𝒰	PROPN
easat-9471	105	9	∪	∪	VERB
easat-9471	105	10	ℐ	ℐ	PROPN
easat-9471	105	11	,	,	PUNCT
easat-9471	105	12	 	 	SPACE
easat-9471	105	13	ℰ	ℰ	NOUN
easat-9471	105	14	)	)	PUNCT
easat-9471	105	15	with	with	ADP
easat-9471	105	16	user	user	NOUN
easat-9471	105	17	set	set	VERB
easat-9471	105	18	𝒰	𝒰	PROPN
easat-9471	105	19	=	=	SYM
easat-9471	105	20	(	(	PUNCT
easat-9471	105	21	𝑢1	𝑢1	PROPN
easat-9471	105	22	,	,	PUNCT
easat-9471	105	23	…	…	PUNCT
easat-9471	105	24	,	,	PUNCT
easat-9471	105	25	𝑢|𝒰|	𝑢|𝒰|	PROPN
easat-9471	105	26	)	)	PUNCT
easat-9471	105	27	and	and	CCONJ
easat-9471	105	28	item	item	NOUN
easat-9471	105	29	set	set	VERB
easat-9471	105	30	ℐ	ℐ	PROPN
easat-9471	105	31	=	=	SYM
easat-9471	105	32	(	(	PUNCT
easat-9471	105	33	𝑖1	𝑖1	PROPN
easat-9471	105	34	,	,	PUNCT
easat-9471	105	35	…	…	PUNCT
easat-9471	105	36	,	,	PUNCT
easat-9471	105	37	𝑖|ℐ|	𝑖|ℐ|	PROPN
easat-9471	105	38	)	)	PUNCT
easat-9471	105	39	the	the	DET
easat-9471	105	40	observed	observe	VERB
easat-9471	105	41	interaction	interaction	NOUN
easat-9471	105	42	set	set	VERB
easat-9471	105	43	ℰ	ℰ	PROPN
easat-9471	105	44	⊆	⊆	NUM
easat-9471	105	45	𝒰	𝒰	PROPN
easat-9471	105	46	×	×	NOUN
easat-9471	105	47	ℐ	ℐ	PRON
easat-9471	105	48	is	be	AUX
easat-9471	105	49	used	use	VERB
easat-9471	105	50	for	for	ADP
easat-9471	105	51	training	training	NOUN
easat-9471	105	52	.	.	PUNCT
easat-9471	106	1	first	first	ADV
easat-9471	106	2	,	,	PUNCT
easat-9471	106	3	we	we	PRON
easat-9471	106	4	define	define	VERB
easat-9471	106	5	an	an	DET
easat-9471	106	6	initial	initial	ADJ
easat-9471	106	7	embedding	embed	VERB
easat-9471	106	8	matrix	matrix	NOUN
easat-9471	106	9	𝑀(0	𝑀(0	NUM
easat-9471	106	10	)	)	PUNCT
easat-9471	106	11	∈	∈	PROPN
easat-9471	106	12	𝑅(|𝕌|+|𝕀|)×𝑑that	𝑅(|𝕌|+|𝕀|)×𝑑that	ADP
easat-9471	106	13	stores	store	NOUN
easat-9471	106	14	trainable	trainable	ADJ
easat-9471	106	15	𝑑-dimensional	𝑑-dimensional	ADJ
easat-9471	106	16	vectors	vector	NOUN
easat-9471	106	17	for	for	ADP
easat-9471	106	18	all	all	DET
easat-9471	106	19	nodes	node	NOUN
easat-9471	106	20	.	.	PUNCT
easat-9471	107	1	we	we	PRON
easat-9471	107	2	then	then	ADV
easat-9471	107	3	construct	construct	VERB
easat-9471	107	4	the	the	DET
easat-9471	107	5	adjacency	adjacency	NOUN
easat-9471	107	6	matrix	matrix	NOUN
easat-9471	107	7	𝐴	𝐴	PROPN
easat-9471	107	8	∈	∈	PROPN
easat-9471	107	9	(	(	PUNCT
easat-9471	107	10	0,1)(|𝒰|+|ℐ|)×(|𝒰|+|ℐ|	0,1)(|𝒰|+|ℐ|)×(|𝒰|+|ℐ|	NUM
easat-9471	107	11	)	)	PUNCT
easat-9471	107	12	,	,	PUNCT
easat-9471	107	13	with	with	ADP
easat-9471	107	14	𝐴𝑢,𝑖	𝐴𝑢,𝑖	PROPN
easat-9471	107	15	=	=	SYM
easat-9471	107	16	1	1	NUM
easat-9471	107	17	if	if	SCONJ
easat-9471	107	18	(	(	PUNCT
easat-9471	107	19	𝑢	𝑢	X
easat-9471	107	20	,	,	PUNCT
easat-9471	107	21	𝑖	𝑖	PART
easat-9471	107	22	)	)	PUNCT
easat-9471	107	23	∈	∈	PROPN
easat-9471	107	24	ℰ	ℰ	PROPN
easat-9471	107	25	,	,	PUNCT
easat-9471	107	26	0	0	NUM
easat-9471	107	27	otherwise	otherwise	ADV
easat-9471	107	28	.	.	PUNCT
easat-9471	108	1	the	the	DET
easat-9471	108	2	adjacency	adjacency	NOUN
easat-9471	108	3	matrix	matrix	NOUN
easat-9471	108	4	is	be	AUX
easat-9471	108	5	normalized	normalize	VERB
easat-9471	108	6	as	as	ADP
easat-9471	108	7	�	�	PROPN
easat-9471	108	8	̃	̃	PROPN
easat-9471	108	9	�	�	PROPN
easat-9471	108	10	=	=	SYM
easat-9471	108	11	𝐷−1/2	𝐷−1/2	PROPN
easat-9471	108	12	 	 	SPACE
easat-9471	108	13	𝐴	𝐴	PROPN
easat-9471	108	14	 	 	SPACE
easat-9471	108	15	𝐷−1/2	𝐷−1/2	PROPN
easat-9471	108	16	where	where	SCONJ
easat-9471	108	17	𝐷	𝐷	PROPN
easat-9471	108	18	is	be	AUX
easat-9471	108	19	the	the	DET
easat-9471	108	20	diagonal	diagonal	ADJ
easat-9471	108	21	degree	degree	NOUN
easat-9471	108	22	matrix	matrix	NOUN
easat-9471	108	23	.	.	PUNCT
easat-9471	109	1	by	by	ADP
easat-9471	109	2	propagating	propagate	VERB
easat-9471	109	3	embeddings	embedding	NOUN
easat-9471	109	4	through	through	ADP
easat-9471	109	5	𝐾	𝐾	PROPN
easat-9471	109	6	layers	layer	NOUN
easat-9471	109	7	,	,	PUNCT
easat-9471	109	8	we	we	PRON
easat-9471	109	9	obtain	obtain	VERB
easat-9471	109	10	(	(	PUNCT
easat-9471	109	11	𝑀(0	𝑀(0	X
easat-9471	109	12	)	)	PUNCT
easat-9471	109	13	,	,	PUNCT
easat-9471	109	14	…	…	PUNCT
easat-9471	109	15	,	,	PUNCT
easat-9471	109	16	𝑀(𝐾	𝑀(𝐾	NOUN
easat-9471	109	17	)	)	PUNCT
easat-9471	109	18	)	)	PUNCT
easat-9471	109	19	.	.	PUNCT
easat-9471	110	1	the	the	DET
easat-9471	110	2	final	final	ADJ
easat-9471	110	3	graph	graph	NOUN
easat-9471	110	4	embedding	embed	VERB
easat-9471	110	5	is	be	AUX
easat-9471	110	6	generated	generate	VERB
easat-9471	110	7	as	as	SCONJ
easat-9471	110	8	follows	follow	VERB
easat-9471	110	9	:	:	PUNCT
easat-9471	110	10	𝑬𝒈𝒓𝒂𝒑𝒉	𝑬𝒈𝒓𝒂𝒑𝒉	PROPN
easat-9471	110	11	=	=	SYM
easat-9471	110	12	∑𝛼𝑘𝑴	∑𝛼𝑘𝑴	X
easat-9471	110	13	(	(	PUNCT
easat-9471	110	14	𝒌	𝒌	NOUN
easat-9471	110	15	)	)	PUNCT
easat-9471	110	16	𝐾	𝐾	NOUN
easat-9471	110	17	𝑘=0	𝑘=0	ADP
easat-9471	110	18	∈	∈	PROPN
easat-9471	110	19	𝑅(|𝕌|+|𝕀|)×𝑑	𝑅(|𝕌|+|𝕀|)×𝑑	PUNCT
easat-9471	110	20	(	(	PUNCT
easat-9471	110	21	2	2	X
easat-9471	110	22	)	)	PUNCT
easat-9471	110	23	the	the	DET
easat-9471	110	24	models	model	NOUN
easat-9471	110	25	that	that	PRON
easat-9471	110	26	generate	generate	VERB
easat-9471	110	27	embeddings	embedding	NOUN
easat-9471	110	28	using	use	VERB
easat-9471	110	29	the	the	DET
easat-9471	110	30	above	above	ADJ
easat-9471	110	31	architecture	architecture	NOUN
easat-9471	110	32	are	be	AUX
easat-9471	110	33	graphsage	graphsage	NOUN
easat-9471	111	1	[	[	X
easat-9471	111	2	30	30	NUM
easat-9471	111	3	]	]	PUNCT
easat-9471	111	4	and	and	CCONJ
easat-9471	111	5	lightgcn	lightgcn	PROPN
easat-9471	111	6	3.1.3	3.1.3	NUM
easat-9471	111	7	.	.	PUNCT
easat-9471	111	8	integrated	integrate	VERB
easat-9471	111	9	collaborative	collaborative	ADJ
easat-9471	111	10	filtering	filtering	NOUN
easat-9471	111	11	model	model	NOUN
easat-9471	111	12	the	the	DET
easat-9471	111	13	integration	integration	NOUN
easat-9471	111	14	of	of	ADP
easat-9471	111	15	sequential	sequential	ADJ
easat-9471	111	16	and	and	CCONJ
easat-9471	111	17	graph	graph	NOUN
easat-9471	111	18	modules	module	NOUN
easat-9471	111	19	is	be	AUX
easat-9471	111	20	achieved	achieve	VERB
easat-9471	111	21	by	by	ADP
easat-9471	111	22	jointly	jointly	ADV
easat-9471	111	23	learning	learn	VERB
easat-9471	111	24	both	both	CCONJ
easat-9471	111	25	the	the	DET
easat-9471	111	26	temporal	temporal	ADJ
easat-9471	111	27	interaction	interaction	NOUN
easat-9471	111	28	sequence	sequence	NOUN
easat-9471	111	29	𝒮𝓊	𝒮𝓊	PROPN
easat-9471	111	30	=	=	PUNCT
easat-9471	111	31	(	(	PUNCT
easat-9471	111	32	𝑖1	𝑖1	PROPN
easat-9471	111	33	𝑢	𝑢	PART
easat-9471	111	34	,	,	PUNCT
easat-9471	111	35	  	  	SPACE
easat-9471	111	36	𝑖2	𝑖2	PROPN
easat-9471	111	37	𝑢	𝑢	PROPN
easat-9471	111	38	,	,	PUNCT
easat-9471	111	39	 	 	SPACE
easat-9471	111	40	…	…	PUNCT
easat-9471	111	41	,	,	PUNCT
easat-9471	111	42	 	 	SPACE
easat-9471	111	43	𝑖|𝒮𝓊|	𝑖|𝒮𝓊|	VERB
easat-9471	111	44	𝑢	𝑢	NOUN
easat-9471	111	45	)	)	PUNCT
easat-9471	111	46	and	and	CCONJ
easat-9471	111	47	the	the	DET
easat-9471	111	48	bipartite	bipartite	PROPN
easat-9471	111	49	graph	graph	NOUN
easat-9471	111	50	𝒢	𝒢	PROPN
easat-9471	111	51	=	=	PUNCT
easat-9471	111	52	(	(	PUNCT
easat-9471	111	53	𝒰	𝒰	PROPN
easat-9471	111	54	∪	∪	VERB
easat-9471	111	55	ℐ	ℐ	PROPN
easat-9471	111	56	,	,	PUNCT
easat-9471	111	57	 	 	SPACE
easat-9471	111	58	ℰ	ℰ	NOUN
easat-9471	111	59	)	)	PUNCT
easat-9471	111	60	in	in	ADP
easat-9471	111	61	a	a	DET
easat-9471	111	62	single	single	ADJ
easat-9471	111	63	embedding	embed	VERB
easat-9471	111	64	space	space	NOUN
easat-9471	111	65	.	.	PUNCT
easat-9471	112	1	to	to	PART
easat-9471	112	2	store	store	VERB
easat-9471	112	3	trainable	trainable	ADJ
easat-9471	112	4	𝑑-dimensional	𝑑-dimensional	ADJ
easat-9471	112	5	vectors	vector	NOUN
easat-9471	112	6	for	for	ADP
easat-9471	112	7	every	every	DET
easat-9471	112	8	node	node	NOUN
easat-9471	112	9	𝑣	𝑣	ADP
easat-9471	112	10	∈	∈	PROPN
easat-9471	112	11	𝒰	𝒰	PROPN
easat-9471	112	12	∪	∪	NOUN
easat-9471	112	13	ℐ	ℐ	PROPN
easat-9471	112	14	,	,	PUNCT
easat-9471	112	15	we	we	PRON
easat-9471	112	16	define	define	VERB
easat-9471	112	17	an	an	DET
easat-9471	112	18	initial	initial	ADJ
easat-9471	112	19	embedding	embed	VERB
easat-9471	112	20	matrix	matrix	NOUN
easat-9471	112	21	𝑀(0	𝑀(0	NUM
easat-9471	112	22	)	)	PUNCT
easat-9471	112	23	∈	∈	PROPN
easat-9471	112	24	𝑅(|𝕌|+|𝕀|)×𝑑	𝑅(|𝕌|+|𝕀|)×𝑑	PUNCT
easat-9471	112	25	.	.	PUNCT
easat-9471	113	1	the	the	DET
easat-9471	113	2	sequential	sequential	ADJ
easat-9471	113	3	encoder	encoder	NOUN
easat-9471	113	4	then	then	ADV
easat-9471	113	5	computes	compute	VERB
easat-9471	113	6	representations	representation	NOUN
easat-9471	113	7	𝑒𝑣	𝑒𝑣	ADP
easat-9471	113	8	𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	ADJ
easat-9471	113	9	and	and	CCONJ
easat-9471	113	10	the	the	DET
easat-9471	113	11	graph	graph	NOUN
easat-9471	113	12	encoder	encoder	NOUN
easat-9471	113	13	computes	compute	VERB
easat-9471	113	14	𝑒𝑣	𝑒𝑣	ADP
easat-9471	113	15	𝑔𝑟𝑎𝑝ℎ	𝑔𝑟𝑎𝑝ℎ	NOUN
easat-9471	113	16	.	.	PUNCT
easat-9471	114	1	by	by	ADP
easat-9471	114	2	applying	apply	VERB
easat-9471	114	3	an	an	DET
easat-9471	114	4	alignment	alignment	NOUN
easat-9471	114	5	and	and	CCONJ
easat-9471	114	6	uniformity	uniformity	NOUN
easat-9471	114	7	loss	loss	NOUN
easat-9471	114	8	over	over	ADP
easat-9471	114	9	the	the	DET
easat-9471	114	10	two	two	NUM
easat-9471	114	11	views	view	NOUN
easat-9471	114	12	of	of	ADP
easat-9471	114	13	each	each	DET
easat-9471	114	14	node	node	NOUN
easat-9471	114	15	(	(	PUNCT
easat-9471	114	16	 	 	SPACE
easat-9471	114	17	𝑒𝑣	𝑒𝑣	ADP
easat-9471	114	18	𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	ADV
easat-9471	114	19	,	,	PUNCT
easat-9471	114	20	 	 	SPACE
easat-9471	114	21	𝑒𝑣	𝑒𝑣	ADP
easat-9471	114	22	𝑔𝑟𝑎𝑝ℎ	𝑔𝑟𝑎𝑝ℎ	PROPN
easat-9471	114	23	 	 	SPACE
easat-9471	114	24	)	)	PUNCT
easat-9471	114	25	,	,	PUNCT
easat-9471	114	26	we	we	PRON
easat-9471	114	27	obtain	obtain	VERB
easat-9471	114	28	the	the	DET
easat-9471	114	29	final	final	ADJ
easat-9471	114	30	integrated	integrate	VERB
easat-9471	114	31	embeddings	embedding	NOUN
easat-9471	114	32	as	as	SCONJ
easat-9471	114	33	follows	follow	VERB
easat-9471	114	34	:	:	PUNCT
easat-9471	114	35	𝐸𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑒𝑑	𝐸𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑒𝑑	PROPN
easat-9471	114	36	=	=	SYM
easat-9471	114	37	β	β	NOUN
easat-9471	114	38	 	 	SPACE
easat-9471	114	39	𝑒𝑣	𝑒𝑣	ADP
easat-9471	114	40	𝑔𝑟𝑎𝑝ℎ	𝑔𝑟𝑎𝑝ℎ	NOUN
easat-9471	114	41	+	+	CCONJ
easat-9471	114	42	(	(	PUNCT
easat-9471	114	43	1	1	NUM
easat-9471	114	44	−	−	NOUN
easat-9471	114	45	β	β	X
easat-9471	114	46	)	)	PUNCT
easat-9471	114	47	 	 	SPACE
easat-9471	114	48	𝑒𝑣	𝑒𝑣	ADP
easat-9471	114	49	𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	ADV
easat-9471	114	50	(	(	PUNCT
easat-9471	114	51	3	3	X
easat-9471	114	52	)	)	PUNCT
easat-9471	114	53	the	the	DET
easat-9471	114	54	model	model	NOUN
easat-9471	114	55	that	that	PRON
easat-9471	114	56	generates	generate	VERB
easat-9471	114	57	embeddings	embedding	NOUN
easat-9471	114	58	using	use	VERB
easat-9471	114	59	the	the	DET
easat-9471	114	60	above	above	ADJ
easat-9471	114	61	architecture	architecture	NOUN
easat-9471	114	62	is	be	AUX
easat-9471	114	63	gsau	gsau	ADJ
easat-9471	114	64	.	.	PROPN
easat-9471	114	65	3.2	3.2	NUM
easat-9471	114	66	.	.	PUNCT
easat-9471	115	1	applying	apply	VERB
easat-9471	115	2	the	the	DET
easat-9471	115	3	grpo	grpo	NOUN
easat-9471	115	4	algorithm	algorithm	NOUN
easat-9471	115	5	in	in	ADP
easat-9471	115	6	recommender	recommender	NOUN
easat-9471	115	7	systems	system	NOUN
easat-9471	115	8	we	we	PRON
easat-9471	115	9	fine	fine	ADJ
easat-9471	115	10	-	-	PUNCT
easat-9471	115	11	tune	tune	NOUN
easat-9471	115	12	the	the	DET
easat-9471	115	13	embeddings	embedding	NOUN
easat-9471	115	14	from	from	ADP
easat-9471	115	15	pre	pre	ADJ
easat-9471	115	16	-	-	ADJ
easat-9471	115	17	trained	train	VERB
easat-9471	115	18	collaborative	collaborative	ADJ
easat-9471	115	19	filtering	filtering	NOUN
easat-9471	115	20	models	model	NOUN
easat-9471	115	21	by	by	ADP
easat-9471	115	22	applying	apply	VERB
easat-9471	115	23	the	the	DET
easat-9471	115	24	grpo	grpo	NOUN
easat-9471	115	25	algorithm	algorithm	NOUN
easat-9471	115	26	.	.	PUNCT
easat-9471	116	1	when	when	SCONJ
easat-9471	116	2	grpo	grpo	NOUN
easat-9471	116	3	is	be	AUX
easat-9471	116	4	applied	apply	VERB
easat-9471	116	5	,	,	PUNCT
easat-9471	116	6	the	the	DET
easat-9471	116	7	sequential	sequential	ADJ
easat-9471	116	8	model	model	NOUN
easat-9471	116	9	state	state	NOUN
easat-9471	116	10	is	be	AUX
easat-9471	116	11	defined	define	VERB
easat-9471	116	12	as	as	ADP
easat-9471	116	13	𝑠𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	𝑠𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	ADJ
easat-9471	116	14	≔	≔	NOUN
easat-9471	116	15	𝑒𝑣	𝑒𝑣	ADP
easat-9471	116	16	𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	𝑠𝑒𝑞𝑢𝑒𝑛𝑡𝑖𝑎𝑙	ADV
easat-9471	116	17	=	=	PUNCT
easat-9471	116	18	(	(	PUNCT
easat-9471	116	19	𝑒𝑖𝑡𝑒𝑚	𝑒𝑖𝑡𝑒𝑚	INTJ
easat-9471	116	20	,	,	PUNCT
easat-9471	116	21	 	 	SPACE
easat-9471	116	22	𝑒𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛	𝑒𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛	NOUN
easat-9471	116	23	)	)	PUNCT
easat-9471	116	24	,	,	PUNCT
easat-9471	116	25	the	the	DET
easat-9471	116	26	graph	graph	NOUN
easat-9471	116	27	model	model	NOUN
easat-9471	116	28	state	state	NOUN
easat-9471	116	29	as	as	SCONJ
easat-9471	116	30	𝑠𝑔𝑟𝑎𝑝ℎ	𝑠𝑔𝑟𝑎𝑝ℎ	X
easat-9471	116	31	≔	≔	VERB
easat-9471	116	32	𝑒𝑣	𝑒𝑣	ADP
easat-9471	116	33	𝑔𝑟𝑎𝑝ℎ	𝑔𝑟𝑎𝑝ℎ	NOUN
easat-9471	116	34	=	=	SYM
easat-9471	116	35	(	(	PUNCT
easat-9471	116	36	𝑒𝑖𝑡𝑒𝑚	𝑒𝑖𝑡𝑒𝑚	INTJ
easat-9471	116	37	,	,	PUNCT
easat-9471	116	38	 	 	SPACE
easat-9471	116	39	𝑒𝑢𝑠𝑒𝑟	𝑒𝑢𝑠𝑒𝑟	PROPN
easat-9471	116	40	)	)	PUNCT
easat-9471	116	41	,	,	PUNCT
easat-9471	116	42	and	and	CCONJ
easat-9471	116	43	the	the	DET
easat-9471	116	44	integrated	integrated	ADJ
easat-9471	116	45	model	model	NOUN
easat-9471	116	46	state	state	NOUN
easat-9471	116	47	as	as	SCONJ
easat-9471	116	48	𝑠𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑒𝑑	𝑠𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑒𝑑	ADV
easat-9471	116	49	≔	≔	NOUN
easat-9471	116	50	𝑒𝑣	𝑒𝑣	ADP
easat-9471	116	51	𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑒𝑑	𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑒𝑑	NOUN
easat-9471	116	52	=	=	SYM
easat-9471	116	53	(	(	PUNCT
easat-9471	116	54	𝑒𝑖𝑡𝑒𝑚	𝑒𝑖𝑡𝑒𝑚	INTJ
easat-9471	116	55	,	,	PUNCT
easat-9471	116	56	 	 	SPACE
easat-9471	116	57	𝑒𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛	𝑒𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛	NOUN
easat-9471	116	58	,	,	PUNCT
easat-9471	116	59	𝑒𝑢𝑠𝑒𝑟	𝑒𝑢𝑠𝑒𝑟	PROPN
easat-9471	116	60	)	)	PUNCT
easat-9471	116	61	.	.	PUNCT
easat-9471	117	1	given	give	VERB
easat-9471	117	2	a	a	DET
easat-9471	117	3	state	state	NOUN
easat-9471	117	4	𝑠	𝑠	NOUN
easat-9471	117	5	,	,	PUNCT
easat-9471	117	6	we	we	PRON
easat-9471	117	7	sample	sample	VERB
easat-9471	117	8	𝐺	𝐺	NOUN
easat-9471	117	9	candidate	candidate	NOUN
easat-9471	117	10	items	item	NOUN
easat-9471	117	11	(	(	PUNCT
easat-9471	117	12	𝑜1	𝑜1	PROPN
easat-9471	117	13	,	,	PUNCT
easat-9471	117	14	…	…	PUNCT
easat-9471	117	15	,	,	PUNCT
easat-9471	117	16	𝑜𝐺	𝑜𝐺	NOUN
easat-9471	117	17	)	)	PUNCT
easat-9471	117	18	from	from	ADP
easat-9471	117	19	the	the	DET
easat-9471	117	20	old	old	ADJ
easat-9471	117	21	policy	policy	NOUN
easat-9471	117	22	πθold	πθold	VERB
easat-9471	117	23	;	;	PUNCT
easat-9471	117	24	for	for	ADP
easat-9471	117	25	each	each	DET
easat-9471	117	26	𝑜	𝑜	NOUN
easat-9471	117	27	,	,	PUNCT
easat-9471	117	28	we	we	PRON
easat-9471	117	29	observe	observe	VERB
easat-9471	117	30	a	a	DET
easat-9471	117	31	verifiable	verifiable	ADJ
easat-9471	117	32	binary	binary	ADJ
easat-9471	117	33	reward	reward	NOUN
easat-9471	117	34	𝑟𝑡	𝑟𝑡	PROPN
easat-9471	117	35	=	=	SYM
easat-9471	117	36	𝑟(𝑠	𝑟(𝑠	PROPN
easat-9471	117	37	,	,	PUNCT
easat-9471	117	38	𝑜	𝑜	NOUN
easat-9471	117	39	)	)	PUNCT
easat-9471	117	40	∈	∈	PROPN
easat-9471	117	41	(	(	PUNCT
easat-9471	117	42	0,1	0,1	NOUN
easat-9471	117	43	)	)	PUNCT
easat-9471	117	44	and	and	CCONJ
easat-9471	117	45	estimate	estimate	VERB
easat-9471	117	46	the	the	DET
easat-9471	117	47	current	current	ADJ
easat-9471	117	48	policy	policy	NOUN
easat-9471	117	49	’s	’s	PART
easat-9471	117	50	success	success	NOUN
easat-9471	117	51	probability	probability	NOUN
easat-9471	117	52	𝑝	𝑝	NOUN
easat-9471	117	53	by	by	ADP
easat-9471	117	54	𝑝	𝑝	PROPN
easat-9471	117	55	=	=	SYM
easat-9471	117	56	1	1	NUM
easat-9471	117	57	𝐺	𝐺	NOUN
easat-9471	117	58	∑	∑	PROPN
easat-9471	117	59	𝑟𝑡	𝑟𝑡	PROPN
easat-9471	117	60	𝐺	𝐺	PROPN
easat-9471	117	61	𝑡=1	𝑡=1	PROPN
easat-9471	117	62	.	.	PUNCT
easat-9471	118	1	we	we	PRON
easat-9471	118	2	then	then	ADV
easat-9471	118	3	define	define	VERB
easat-9471	118	4	the	the	DET
easat-9471	118	5	grpo	grpo	NOUN
easat-9471	118	6	advantage	advantage	NOUN
easat-9471	118	7	estimate	estimate	NOUN
easat-9471	118	8	𝐴(𝑠	𝐴(𝑠	NOUN
easat-9471	118	9	,	,	PUNCT
easat-9471	118	10	𝑜	𝑜	NOUN
easat-9471	118	11	)	)	PUNCT
easat-9471	118	12	as	as	SCONJ
easat-9471	118	13	follows	follow	VERB
easat-9471	118	14	:	:	PUNCT
easat-9471	118	15	876	876	NUM
easat-9471	118	16	edelweiss	edelweiss	PROPN
easat-9471	118	17	applied	apply	VERB
easat-9471	118	18	science	science	NOUN
easat-9471	118	19	and	and	CCONJ
easat-9471	118	20	technology	technology	NOUN
easat-9471	118	21	issn	issn	PROPN
easat-9471	118	22	:	:	PUNCT
easat-9471	118	23	2576	2576	NUM
easat-9471	118	24	-	-	SYM
easat-9471	118	25	8484	8484	NUM
easat-9471	118	26	vol	vol	NOUN
easat-9471	118	27	.	.	PROPN
easat-9471	119	1	9	9	NUM
easat-9471	119	2	,	,	PUNCT
easat-9471	119	3	no	no	INTJ
easat-9471	119	4	.	.	NOUN
easat-9471	119	5	8	8	NUM
easat-9471	119	6	:	:	PUNCT
easat-9471	119	7	871	871	NUM
easat-9471	119	8	-	-	SYM
easat-9471	119	9	881	881	NUM
easat-9471	119	10	,	,	PUNCT
easat-9471	119	11	2025	2025	NUM
easat-9471	119	12	doi	doi	NOUN
easat-9471	119	13	:	:	PUNCT
easat-9471	119	14	10.55214/2576	10.55214/2576	NUM
easat-9471	119	15	-	-	PUNCT
easat-9471	119	16	8484.v9i8.9471	8484.v9i8.9471	NOUN
easat-9471	120	1	©	©	PROPN
easat-9471	120	2	2025	2025	NUM
easat-9471	120	3	by	by	ADP
easat-9471	120	4	the	the	DET
easat-9471	120	5	authors	author	NOUN
easat-9471	120	6	;	;	PUNCT
easat-9471	120	7	licensee	licensee	PROPN
easat-9471	120	8	learning	learn	VERB
easat-9471	120	9	gate	gate	PROPN
easat-9471	120	10	𝐴(𝑠	𝐴(𝑠	PROPN
easat-9471	120	11	,	,	PUNCT
easat-9471	120	12	𝑜	𝑜	NOUN
easat-9471	120	13	)	)	PUNCT
easat-9471	120	14	=	=	PRON
easat-9471	120	15	{	{	PUNCT
easat-9471	120	16	√	√	NUM
easat-9471	120	17	1	1	NUM
easat-9471	120	18	−	−	PROPN
easat-9471	120	19	𝑝	𝑝	PROPN
easat-9471	120	20	𝑝	𝑝	PROPN
easat-9471	121	1	if	if	SCONJ
easat-9471	121	2	𝑟(𝑠	𝑟(𝑠	PROPN
easat-9471	121	3	,	,	PUNCT
easat-9471	121	4	𝑜	𝑜	NOUN
easat-9471	121	5	)	)	PUNCT
easat-9471	121	6	=	=	SYM
easat-9471	121	7	1	1	NUM
easat-9471	121	8	−√	−√	NOUN
easat-9471	121	9	𝑝	𝑝	NOUN
easat-9471	121	10	1	1	NUM
easat-9471	121	11	−	−	PROPN
easat-9471	121	12	𝑝	𝑝	NOUN
easat-9471	121	13	if	if	SCONJ
easat-9471	121	14	𝑟(𝑠	𝑟(𝑠	PROPN
easat-9471	121	15	,	,	PUNCT
easat-9471	121	16	𝑜	𝑜	NOUN
easat-9471	121	17	)	)	PUNCT
easat-9471	121	18	=	=	SYM
easat-9471	121	19	0	0	PUNCT
easat-9471	121	20	(	(	PUNCT
easat-9471	121	21	4	4	NUM
easat-9471	121	22	)	)	PUNCT
easat-9471	121	23	next	next	ADV
easat-9471	121	24	,	,	PUNCT
easat-9471	121	25	we	we	PRON
easat-9471	121	26	compute	compute	VERB
easat-9471	121	27	the	the	DET
easat-9471	121	28	importance	importance	NOUN
easat-9471	121	29	ratio	ratio	NOUN
easat-9471	121	30	ρ𝑡	ρ𝑡	NOUN
easat-9471	121	31	=	=	SYM
easat-9471	121	32	πθnew(𝑜𝑡∣∣𝑠	πθnew(𝑜𝑡∣∣𝑠	PROPN
easat-9471	121	33	)	)	PUNCT
easat-9471	121	34	πθold	πθold	PROPN
easat-9471	121	35	(	(	PUNCT
easat-9471	121	36	𝑜𝑡∣∣𝑠	𝑜𝑡∣∣𝑠	PROPN
easat-9471	121	37	)	)	PUNCT
easat-9471	121	38	,	,	PUNCT
easat-9471	121	39	and	and	CCONJ
easat-9471	121	40	apply	apply	VERB
easat-9471	121	41	the	the	DET
easat-9471	121	42	ppo	ppo	PROPN
easat-9471	121	43	clipping	clip	VERB
easat-9471	121	44	to	to	ADP
easat-9471	121	45	each	each	DET
easat-9471	121	46	candidate	candidate	NOUN
easat-9471	121	47	𝑜𝑡	𝑜𝑡	VERB
easat-9471	121	48	to	to	PART
easat-9471	121	49	obtain	obtain	VERB
easat-9471	121	50	gradient	gradient	ADJ
easat-9471	121	51	optimization	optimization	NOUN
easat-9471	121	52	objective	objective	NOUN
easat-9471	121	53	as	as	SCONJ
easat-9471	121	54	follows	follow	VERB
easat-9471	121	55	:	:	PUNCT
easat-9471	121	56	𝐽grpo(θ	𝐽grpo(θ	X
easat-9471	121	57	)	)	PUNCT
easat-9471	121	58	=	=	SYM
easat-9471	121	59	1	1	NUM
easat-9471	121	60	𝐺	𝐺	PROPN
easat-9471	121	61	∑[min(ρ𝑡𝐴(𝑠	∑[min(ρ𝑡𝐴(𝑠	NOUN
easat-9471	121	62	,	,	PUNCT
easat-9471	121	63	𝑜𝑡	𝑜𝑡	PROPN
easat-9471	121	64	)	)	PUNCT
easat-9471	121	65	,	,	PUNCT
easat-9471	121	66	 	 	SPACE
easat-9471	121	67	clip(ρ𝑡	clip(ρ𝑡	NOUN
easat-9471	121	68	,	,	PUNCT
easat-9471	121	69	1	1	NUM
easat-9471	121	70	−	−	NOUN
easat-9471	121	71	ε	ε	PROPN
easat-9471	121	72	,	,	PUNCT
easat-9471	121	73	1	1	NUM
easat-9471	121	74	+	+	CCONJ
easat-9471	121	75	ε	ε	PROPN
easat-9471	121	76	)	)	PUNCT
easat-9471	121	77	 	 	SPACE
easat-9471	121	78	𝐴(𝑠	𝐴(𝑠	PROPN
easat-9471	121	79	,	,	PUNCT
easat-9471	121	80	𝑜𝑡	𝑜𝑡	NOUN
easat-9471	121	81	)	)	PUNCT
easat-9471	121	82	)	)	PUNCT
easat-9471	121	83	−	−	PROPN
easat-9471	122	1	β	β	X
easat-9471	122	2	 	 	SPACE
easat-9471	122	3	𝐷kl(πθold	𝐷kl(πθold	PROPN
easat-9471	122	4	|πref	|πref	PROPN
easat-9471	122	5	)	)	PUNCT
easat-9471	122	6	]	]	PUNCT
easat-9471	123	1	𝐺	𝐺	PROPN
easat-9471	123	2	𝑡=1	𝑡=1	PROPN
easat-9471	123	3	  	  	SPACE
easat-9471	123	4	(	(	PUNCT
easat-9471	123	5	5	5	NUM
easat-9471	123	6	)	)	PUNCT
easat-9471	123	7	here	here	ADV
easat-9471	123	8	,	,	PUNCT
easat-9471	123	9	ε	ε	PROPN
easat-9471	123	10	∈	∈	PROPN
easat-9471	123	11	(	(	PUNCT
easat-9471	123	12	0,1	0,1	NOUN
easat-9471	123	13	)	)	PUNCT
easat-9471	123	14	is	be	AUX
easat-9471	123	15	a	a	DET
easat-9471	123	16	hyperparameter	hyperparameter	NOUN
easat-9471	123	17	that	that	PRON
easat-9471	123	18	bounds	bound	VERB
easat-9471	123	19	the	the	DET
easat-9471	123	20	importance	importance	NOUN
easat-9471	123	21	ratio	ratio	NOUN
easat-9471	123	22	within	within	ADP
easat-9471	123	23	[	[	X
easat-9471	123	24	1	1	NUM
easat-9471	123	25	−	−	PROPN
easat-9471	123	26	ε	ε	PROPN
easat-9471	123	27	,	,	PUNCT
easat-9471	123	28	 	 	SPACE
easat-9471	123	29	1	1	NUM
easat-9471	123	30	+	+	CCONJ
easat-9471	123	31	ε	ε	PROPN
easat-9471	123	32	]	]	X
easat-9471	123	33	,	,	PUNCT
easat-9471	123	34	and	and	CCONJ
easat-9471	123	35	β	β	X
easat-9471	123	36	>	>	X
easat-9471	123	37	0	0	NUM
easat-9471	123	38	is	be	AUX
easat-9471	123	39	the	the	DET
easat-9471	123	40	kl	kl	NOUN
easat-9471	123	41	-	-	ADJ
easat-9471	123	42	regularization	regularization	NOUN
easat-9471	123	43	coefficient	coefficient	NOUN
easat-9471	123	44	controlling	control	VERB
easat-9471	123	45	the	the	DET
easat-9471	123	46	divergence	divergence	NOUN
easat-9471	123	47	between	between	ADP
easat-9471	123	48	the	the	DET
easat-9471	123	49	new	new	ADJ
easat-9471	123	50	policy	policy	NOUN
easat-9471	123	51	and	and	CCONJ
easat-9471	123	52	the	the	DET
easat-9471	123	53	reference	reference	NOUN
easat-9471	123	54	policy	policy	NOUN
easat-9471	123	55	.	.	PUNCT
easat-9471	124	1	through	through	ADP
easat-9471	124	2	this	this	DET
easat-9471	124	3	process	process	NOUN
easat-9471	124	4	,	,	PUNCT
easat-9471	124	5	grpo	grpo	NOUN
easat-9471	124	6	repeatedly	repeatedly	ADV
easat-9471	124	7	amplifies	amplify	VERB
easat-9471	124	8	the	the	DET
easat-9471	124	9	reference	reference	NOUN
easat-9471	124	10	policy	policy	NOUN
easat-9471	124	11	’s	’s	PART
easat-9471	124	12	success	success	NOUN
easat-9471	124	13	probability	probability	NOUN
easat-9471	124	14	𝑝	𝑝	NOUN
easat-9471	124	15	and	and	CCONJ
easat-9471	124	16	enhances	enhance	VERB
easat-9471	124	17	convergence	convergence	NOUN
easat-9471	124	18	stability	stability	NOUN
easat-9471	124	19	during	during	ADP
easat-9471	124	20	training	training	NOUN
easat-9471	124	21	,	,	PUNCT
easat-9471	124	22	thereby	thereby	ADV
easat-9471	124	23	achieving	achieve	VERB
easat-9471	124	24	model	model	NOUN
easat-9471	124	25	optimization	optimization	NOUN
easat-9471	124	26	[	[	X
easat-9471	124	27	31	31	NUM
easat-9471	124	28	]	]	PUNCT
easat-9471	124	29	.	.	PUNCT
easat-9471	125	1	4	4	X
easat-9471	125	2	.	.	X
easat-9471	125	3	results	result	VERB
easat-9471	125	4	4.1	4.1	NUM
easat-9471	125	5	.	.	PUNCT
easat-9471	126	1	experimental	experimental	ADJ
easat-9471	126	2	setup	setup	NOUN
easat-9471	126	3	4.1.1	4.1.1	NUM
easat-9471	126	4	.	.	PUNCT
easat-9471	127	1	datasets	dataset	NOUN
easat-9471	127	2	for	for	ADP
easat-9471	127	3	model	model	NOUN
easat-9471	127	4	evaluation	evaluation	NOUN
easat-9471	127	5	,	,	PUNCT
easat-9471	127	6	we	we	PRON
easat-9471	127	7	utilized	utilize	VERB
easat-9471	127	8	datasets	dataset	NOUN
easat-9471	127	9	from	from	ADP
easat-9471	127	10	amazon	amazon	NOUN
easat-9471	127	11	recommendation	recommendation	NOUN
easat-9471	127	12	reviews	review	NOUN
easat-9471	127	13	hou	hou	PROPN
easat-9471	127	14	,	,	PUNCT
easat-9471	127	15	et	et	PROPN
easat-9471	127	16	al	al	PROPN
easat-9471	127	17	.	.	PUNCT
easat-9471	128	1	[	[	X
easat-9471	128	2	32	32	NUM
easat-9471	128	3	]	]	PUNCT
easat-9471	128	4	in	in	ADP
easat-9471	128	5	the	the	DET
easat-9471	128	6	fields	field	NOUN
easat-9471	128	7	of	of	ADP
easat-9471	128	8	baby	baby	NOUN
easat-9471	128	9	products	product	NOUN
easat-9471	128	10	,	,	PUNCT
easat-9471	128	11	video	video	NOUN
easat-9471	128	12	games	game	NOUN
easat-9471	128	13	,	,	PUNCT
easat-9471	128	14	and	and	CCONJ
easat-9471	128	15	industrial	industrial	ADJ
easat-9471	128	16	and	and	CCONJ
easat-9471	128	17	scientific	scientific	ADJ
easat-9471	128	18	.	.	PUNCT
easat-9471	129	1	we	we	PRON
easat-9471	129	2	intentionally	intentionally	ADV
easat-9471	129	3	selected	select	VERB
easat-9471	129	4	datasets	dataset	NOUN
easat-9471	129	5	with	with	ADP
easat-9471	129	6	varying	vary	VERB
easat-9471	129	7	statistics	statistic	NOUN
easat-9471	129	8	regarding	regard	VERB
easat-9471	129	9	the	the	DET
easat-9471	129	10	number	number	NOUN
easat-9471	129	11	of	of	ADP
easat-9471	129	12	users	user	NOUN
easat-9471	129	13	and	and	CCONJ
easat-9471	129	14	items	item	NOUN
easat-9471	129	15	to	to	PART
easat-9471	129	16	comprehensively	comprehensively	ADV
easat-9471	129	17	evaluate	evaluate	VERB
easat-9471	129	18	the	the	DET
easat-9471	129	19	models	model	NOUN
easat-9471	129	20	’	'	PUNCT
easat-9471	129	21	performance	performance	NOUN
easat-9471	129	22	.	.	PUNCT
easat-9471	130	1	table	table	NOUN
easat-9471	130	2	1	1	NUM
easat-9471	130	3	represents	represent	VERB
easat-9471	130	4	the	the	DET
easat-9471	130	5	statistics	statistic	NOUN
easat-9471	130	6	of	of	ADP
easat-9471	130	7	preprocessed	preprocesse	VERB
easat-9471	130	8	datasets	dataset	NOUN
easat-9471	130	9	.	.	PUNCT
easat-9471	131	1	in	in	ADP
easat-9471	131	2	previous	previous	ADJ
easat-9471	131	3	studies	study	NOUN
easat-9471	131	4	[	[	X
easat-9471	131	5	3	3	NUM
easat-9471	131	6	,	,	PUNCT
easat-9471	131	7	5	5	NUM
easat-9471	131	8	]	]	PUNCT
easat-9471	131	9	as	as	ADP
easat-9471	131	10	part	part	NOUN
easat-9471	131	11	of	of	ADP
easat-9471	131	12	core	core	NOUN
easat-9471	131	13	data	datum	NOUN
easat-9471	131	14	filtering	filtering	NOUN
easat-9471	131	15	,	,	PUNCT
easat-9471	131	16	users	user	NOUN
easat-9471	131	17	and	and	CCONJ
easat-9471	131	18	items	item	NOUN
easat-9471	131	19	with	with	ADP
easat-9471	131	20	fewer	few	ADJ
easat-9471	131	21	than	than	ADP
easat-9471	131	22	interactions	interaction	NOUN
easat-9471	131	23	were	be	AUX
easat-9471	131	24	removed	remove	VERB
easat-9471	131	25	only	only	ADV
easat-9471	131	26	once	once	ADV
easat-9471	131	27	.	.	PUNCT
easat-9471	132	1	however	however	ADV
easat-9471	132	2	,	,	PUNCT
easat-9471	132	3	in	in	ADP
easat-9471	132	4	recent	recent	ADJ
easat-9471	132	5	studies	study	NOUN
easat-9471	132	6	[	[	X
easat-9471	132	7	8	8	NUM
easat-9471	132	8	,	,	PUNCT
easat-9471	132	9	33	33	NUM
easat-9471	132	10	]	]	PUNCT
easat-9471	132	11	the	the	DET
easat-9471	132	12	core	core	NOUN
easat-9471	132	13	data	datum	NOUN
easat-9471	132	14	filtering	filtering	NOUN
easat-9471	132	15	method	method	NOUN
easat-9471	132	16	is	be	AUX
easat-9471	132	17	repeatedly	repeatedly	ADV
easat-9471	132	18	executed	execute	VERB
easat-9471	132	19	until	until	SCONJ
easat-9471	132	20	there	there	PRON
easat-9471	132	21	are	be	VERB
easat-9471	132	22	no	no	PRON
easat-9471	132	23	more	more	ADJ
easat-9471	132	24	items	item	NOUN
easat-9471	132	25	to	to	PART
easat-9471	132	26	remove	remove	VERB
easat-9471	132	27	.	.	PUNCT
easat-9471	133	1	therefore	therefore	ADV
easat-9471	133	2	,	,	PUNCT
easat-9471	133	3	in	in	ADP
easat-9471	133	4	this	this	DET
easat-9471	133	5	study	study	NOUN
easat-9471	133	6	,	,	PUNCT
easat-9471	133	7	users	user	NOUN
easat-9471	133	8	and	and	CCONJ
easat-9471	133	9	items	item	NOUN
easat-9471	133	10	with	with	ADP
easat-9471	133	11	fewer	few	ADJ
easat-9471	133	12	than	than	ADP
easat-9471	133	13	five	five	NUM
easat-9471	133	14	interactions	interaction	NOUN
easat-9471	133	15	were	be	AUX
easat-9471	133	16	repeatedly	repeatedly	ADV
easat-9471	133	17	removed	remove	VERB
easat-9471	133	18	to	to	PART
easat-9471	133	19	preprocess	preprocess	VERB
easat-9471	133	20	the	the	DET
easat-9471	133	21	data	datum	NOUN
easat-9471	133	22	.	.	PUNCT
easat-9471	134	1	table	table	NOUN
easat-9471	134	2	1	1	NUM
easat-9471	134	3	.	.	PUNCT
easat-9471	135	1	statistics	statistic	NOUN
easat-9471	135	2	of	of	ADP
easat-9471	135	3	the	the	DET
easat-9471	135	4	dataset	dataset	NOUN
easat-9471	135	5	after	after	ADP
easat-9471	135	6	preprocessing	preprocesse	VERB
easat-9471	135	7	.	.	PUNCT
easat-9471	136	1	avg	avg	PROPN
easat-9471	136	2	.	.	PUNCT
easat-9471	136	3	len	len	PROPN
easat-9471	136	4	denotes	denote	VERB
easat-9471	136	5	the	the	DET
easat-9471	136	6	average	average	ADJ
easat-9471	136	7	sequence	sequence	NOUN
easat-9471	136	8	length	length	NOUN
easat-9471	136	9	of	of	ADP
easat-9471	136	10	users	user	NOUN
easat-9471	136	11	.	.	PUNCT
easat-9471	137	1	datasets	dataset	NOUN
easat-9471	137	2	#	#	SYM
easat-9471	137	3	users	user	NOUN
easat-9471	137	4	#	#	NOUN
easat-9471	137	5	items	item	NOUN
easat-9471	137	6	interactions	interaction	NOUN
easat-9471	137	7	avg	avg	PROPN
easat-9471	137	8	.	.	PUNCT
easat-9471	138	1	len	len	ADJ
easat-9471	138	2	baby	baby	NOUN
easat-9471	138	3	products	product	VERB
easat-9471	138	4	155.434	155.434	NUM
easat-9471	138	5	36.845	36.845	NUM
easat-9471	138	6	1.287.653	1.287.653	NUM
easat-9471	138	7	8.28	8.28	NUM
easat-9471	138	8	video	video	NOUN
easat-9471	138	9	games	game	NOUN
easat-9471	138	10	98.906	98.906	NUM
easat-9471	138	11	26.354	26.354	NUM
easat-9471	138	12	857.505	857.505	NUM
easat-9471	138	13	8.67	8.67	NUM
easat-9471	138	14	industrial	industrial	ADJ
easat-9471	138	15	&	&	CCONJ
easat-9471	138	16	scientific	scientific	ADJ
easat-9471	138	17	54.567	54.567	NUM
easat-9471	138	18	27.229	27.229	NUM
easat-9471	138	19	446.046	446.046	NUM
easat-9471	138	20	8.17	8.17	NUM
easat-9471	138	21	4.1.2	4.1.2	NUM
easat-9471	138	22	.	.	PUNCT
easat-9471	138	23	baseline	baseline	PROPN
easat-9471	138	24	in	in	ADP
easat-9471	138	25	this	this	DET
easat-9471	138	26	study	study	NOUN
easat-9471	138	27	.	.	PUNCT
easat-9471	139	1	the	the	DET
easat-9471	139	2	grpo	grpo	NOUN
easat-9471	139	3	algorithm	algorithm	NOUN
easat-9471	139	4	was	be	AUX
easat-9471	139	5	applied	apply	VERB
easat-9471	139	6	to	to	ADP
easat-9471	139	7	three	three	NUM
easat-9471	139	8	baseline	baseline	ADJ
easat-9471	139	9	models	model	NOUN
easat-9471	139	10	.	.	PUNCT
easat-9471	140	1	among	among	ADP
easat-9471	140	2	sequential	sequential	ADJ
easat-9471	140	3	collaborative	collaborative	ADJ
easat-9471	140	4	filtering	filtering	NOUN
easat-9471	140	5	models	model	NOUN
easat-9471	140	6	.	.	PUNCT
easat-9471	141	1	sasrec	sasrec	NOUN
easat-9471	141	2	was	be	AUX
easat-9471	141	3	used	use	VERB
easat-9471	141	4	as	as	ADP
easat-9471	141	5	the	the	DET
easat-9471	141	6	baseline	baseline	NOUN
easat-9471	141	7	model	model	NOUN
easat-9471	141	8	;	;	PUNCT
easat-9471	141	9	among	among	ADP
easat-9471	141	10	graph	graph	NOUN
easat-9471	141	11	-	-	PUNCT
easat-9471	141	12	based	base	VERB
easat-9471	141	13	collaborative	collaborative	ADJ
easat-9471	141	14	filtering	filtering	NOUN
easat-9471	141	15	models	model	NOUN
easat-9471	141	16	.	.	PUNCT
easat-9471	142	1	lightgcn	lightgcn	PROPN
easat-9471	142	2	was	be	AUX
easat-9471	142	3	used	use	VERB
easat-9471	142	4	;	;	PUNCT
easat-9471	142	5	and	and	CCONJ
easat-9471	142	6	among	among	ADP
easat-9471	142	7	sequential	sequential	ADJ
easat-9471	142	8	and	and	CCONJ
easat-9471	142	9	graph	graph	NOUN
easat-9471	142	10	integrated	integrate	VERB
easat-9471	142	11	collaborative	collaborative	ADJ
easat-9471	142	12	filtering	filtering	NOUN
easat-9471	142	13	models	model	NOUN
easat-9471	142	14	.	.	PUNCT
easat-9471	143	1	gsau	gsau	PROPN
easat-9471	143	2	was	be	AUX
easat-9471	143	3	used	use	VERB
easat-9471	143	4	.	.	PUNCT
easat-9471	144	1	4.1.3	4.1.3	X
easat-9471	144	2	.	.	PUNCT
easat-9471	144	3	evaluation	evaluation	NOUN
easat-9471	144	4	setting	set	VERB
easat-9471	144	5	we	we	PRON
easat-9471	144	6	grouped	group	VERB
easat-9471	144	7	all	all	DET
easat-9471	144	8	interaction	interaction	NOUN
easat-9471	144	9	data	datum	NOUN
easat-9471	144	10	by	by	ADP
easat-9471	144	11	user	user	NOUN
easat-9471	144	12	and	and	CCONJ
easat-9471	144	13	sorted	sort	VERB
easat-9471	144	14	each	each	DET
easat-9471	144	15	user	user	NOUN
easat-9471	144	16	's	's	PART
easat-9471	144	17	interaction	interaction	NOUN
easat-9471	144	18	history	history	NOUN
easat-9471	144	19	chronologically	chronologically	ADV
easat-9471	144	20	.	.	PUNCT
easat-9471	145	1	the	the	DET
easat-9471	145	2	last	last	ADJ
easat-9471	145	3	interaction	interaction	NOUN
easat-9471	145	4	data	datum	NOUN
easat-9471	145	5	in	in	ADP
easat-9471	145	6	the	the	DET
easat-9471	145	7	sorted	sorted	ADJ
easat-9471	145	8	data	datum	NOUN
easat-9471	145	9	was	be	AUX
easat-9471	145	10	assigned	assign	VERB
easat-9471	145	11	to	to	ADP
easat-9471	145	12	the	the	DET
easat-9471	145	13	evaluation	evaluation	NOUN
easat-9471	145	14	set	set	NOUN
easat-9471	145	15	.	.	PUNCT
easat-9471	146	1	the	the	DET
easat-9471	146	2	previous	previous	ADJ
easat-9471	146	3	interaction	interaction	NOUN
easat-9471	146	4	data	datum	NOUN
easat-9471	146	5	was	be	AUX
easat-9471	146	6	assigned	assign	VERB
easat-9471	146	7	to	to	ADP
easat-9471	146	8	the	the	DET
easat-9471	146	9	validation	validation	NOUN
easat-9471	146	10	set	set	NOUN
easat-9471	146	11	.	.	PUNCT
easat-9471	147	1	and	and	CCONJ
easat-9471	147	2	the	the	DET
easat-9471	147	3	rest	rest	NOUN
easat-9471	147	4	was	be	AUX
easat-9471	147	5	used	use	VERB
easat-9471	147	6	as	as	ADP
easat-9471	147	7	the	the	DET
easat-9471	147	8	training	training	NOUN
easat-9471	147	9	set	set	NOUN
easat-9471	147	10	.	.	PUNCT
easat-9471	148	1	in	in	ADP
easat-9471	148	2	the	the	DET
easat-9471	148	3	evaluation	evaluation	NOUN
easat-9471	148	4	877	877	NUM
easat-9471	148	5	edelweiss	edelweiss	PROPN
easat-9471	148	6	applied	apply	VERB
easat-9471	148	7	science	science	NOUN
easat-9471	148	8	and	and	CCONJ
easat-9471	148	9	technology	technology	NOUN
easat-9471	148	10	issn	issn	PROPN
easat-9471	148	11	:	:	PUNCT
easat-9471	148	12	2576	2576	NUM
easat-9471	148	13	-	-	SYM
easat-9471	148	14	8484	8484	NUM
easat-9471	148	15	vol	vol	NOUN
easat-9471	148	16	.	.	PROPN
easat-9471	149	1	9	9	NUM
easat-9471	149	2	,	,	PUNCT
easat-9471	149	3	no	no	INTJ
easat-9471	149	4	.	.	NOUN
easat-9471	149	5	8	8	NUM
easat-9471	149	6	:	:	PUNCT
easat-9471	149	7	871	871	NUM
easat-9471	149	8	-	-	SYM
easat-9471	149	9	881	881	NUM
easat-9471	149	10	,	,	PUNCT
easat-9471	149	11	2025	2025	NUM
easat-9471	149	12	doi	doi	NOUN
easat-9471	149	13	:	:	PUNCT
easat-9471	149	14	10.55214/2576	10.55214/2576	NUM
easat-9471	149	15	-	-	PUNCT
easat-9471	149	16	8484.v9i8.9471	8484.v9i8.9471	NOUN
easat-9471	150	1	©	©	PROPN
easat-9471	150	2	2025	2025	NUM
easat-9471	150	3	by	by	ADP
easat-9471	150	4	the	the	DET
easat-9471	150	5	authors	author	NOUN
easat-9471	150	6	;	;	PUNCT
easat-9471	150	7	licensee	licensee	PROPN
easat-9471	150	8	learning	learning	NOUN
easat-9471	150	9	gate	gate	NOUN
easat-9471	150	10	phase	phase	NOUN
easat-9471	150	11	.	.	PUNCT
easat-9471	151	1	instead	instead	ADV
easat-9471	151	2	of	of	ADP
easat-9471	151	3	the	the	DET
easat-9471	151	4	sampling	sampling	NOUN
easat-9471	151	5	method	method	NOUN
easat-9471	151	6	used	use	VERB
easat-9471	151	7	in	in	ADP
easat-9471	151	8	previous	previous	ADJ
easat-9471	151	9	studies	study	NOUN
easat-9471	151	10	.	.	PUNCT
easat-9471	152	1	we	we	PRON
easat-9471	152	2	performed	perform	VERB
easat-9471	152	3	a	a	DET
easat-9471	152	4	full	full	ADV
easat-9471	152	5	-	-	PUNCT
easat-9471	152	6	ranking	rank	VERB
easat-9471	152	7	evaluation	evaluation	NOUN
easat-9471	152	8	[	[	X
easat-9471	152	9	33	33	NUM
easat-9471	152	10	,	,	PUNCT
easat-9471	152	11	34	34	NUM
easat-9471	152	12	]	]	PUNCT
easat-9471	152	13	with	with	ADP
easat-9471	152	14	all	all	DET
easat-9471	152	15	items	item	NOUN
easat-9471	152	16	as	as	ADP
easat-9471	152	17	candidates	candidate	NOUN
easat-9471	152	18	.	.	PUNCT
easat-9471	153	1	additionally	additionally	ADV
easat-9471	153	2	.	.	PUNCT
easat-9471	154	1	we	we	PRON
easat-9471	154	2	masked	mask	VERB
easat-9471	154	3	past	past	ADJ
easat-9471	154	4	items	item	NOUN
easat-9471	154	5	that	that	PRON
easat-9471	154	6	had	have	AUX
easat-9471	154	7	already	already	ADV
easat-9471	154	8	been	be	AUX
easat-9471	154	9	exposed	expose	VERB
easat-9471	154	10	during	during	ADP
easat-9471	154	11	the	the	DET
easat-9471	154	12	training	training	NOUN
easat-9471	154	13	phase	phase	NOUN
easat-9471	154	14	to	to	PART
easat-9471	154	15	exclude	exclude	VERB
easat-9471	154	16	them	they	PRON
easat-9471	154	17	from	from	ADP
easat-9471	154	18	the	the	DET
easat-9471	154	19	evaluation	evaluation	NOUN
easat-9471	154	20	.	.	PUNCT
easat-9471	155	1	for	for	ADP
easat-9471	155	2	quantitative	quantitative	ADJ
easat-9471	155	3	comparison	comparison	NOUN
easat-9471	155	4	and	and	CCONJ
easat-9471	155	5	evaluation	evaluation	NOUN
easat-9471	155	6	.	.	PUNCT
easat-9471	156	1	we	we	PRON
easat-9471	156	2	used	use	VERB
easat-9471	156	3	recall	recall	NOUN
easat-9471	156	4	and	and	CCONJ
easat-9471	156	5	normalized	normalize	VERB
easat-9471	156	6	discounted	discount	VERB
easat-9471	156	7	cumulative	cumulative	ADJ
easat-9471	156	8	gain	gain	NOUN
easat-9471	156	9	(	(	PUNCT
easat-9471	156	10	ndcg	ndcg	NOUN
easat-9471	156	11	)	)	PUNCT
easat-9471	156	12	scores	score	NOUN
easat-9471	156	13	.	.	PUNCT
easat-9471	157	1	which	which	PRON
easat-9471	157	2	are	be	AUX
easat-9471	157	3	widely	widely	ADV
easat-9471	157	4	used	use	VERB
easat-9471	157	5	as	as	ADP
easat-9471	157	6	evaluation	evaluation	NOUN
easat-9471	157	7	metrics	metric	NOUN
easat-9471	157	8	in	in	ADP
easat-9471	157	9	recommender	recommender	NOUN
easat-9471	157	10	systems	system	NOUN
easat-9471	157	11	.	.	PUNCT
easat-9471	158	1	4.1.4	4.1.4	X
easat-9471	158	2	.	.	PUNCT
easat-9471	159	1	implementation	implementation	NOUN
easat-9471	159	2	details	detail	NOUN
easat-9471	159	3	the	the	DET
easat-9471	159	4	hyperparameters	hyperparameter	NOUN
easat-9471	159	5	of	of	ADP
easat-9471	159	6	the	the	DET
easat-9471	159	7	three	three	NUM
easat-9471	159	8	baseline	baseline	ADJ
easat-9471	159	9	collaborative	collaborative	ADJ
easat-9471	159	10	filtering	filtering	NOUN
easat-9471	159	11	models	model	NOUN
easat-9471	159	12	were	be	AUX
easat-9471	159	13	set	set	VERB
easat-9471	159	14	identically	identically	ADV
easat-9471	159	15	for	for	ADP
easat-9471	159	16	fair	fair	ADJ
easat-9471	159	17	comparison	comparison	NOUN
easat-9471	159	18	.	.	PUNCT
easat-9471	160	1	the	the	DET
easat-9471	160	2	batch	batch	NOUN
easat-9471	160	3	size	size	NOUN
easat-9471	160	4	is	be	AUX
easat-9471	160	5	2048	2048	NUM
easat-9471	160	6	.	.	PUNCT
easat-9471	161	1	the	the	DET
easat-9471	161	2	random	random	ADJ
easat-9471	161	3	seed	seed	NOUN
easat-9471	161	4	is	be	AUX
easat-9471	161	5	2020	2020	NUM
easat-9471	161	6	.	.	PUNCT
easat-9471	162	1	and	and	CCONJ
easat-9471	162	2	the	the	DET
easat-9471	162	3	epoch	epoch	NOUN
easat-9471	162	4	is	be	AUX
easat-9471	162	5	200	200	NUM
easat-9471	162	6	.	.	PUNCT
easat-9471	163	1	for	for	ADP
easat-9471	163	2	grpo	grpo	NOUN
easat-9471	163	3	learning	learning	NOUN
easat-9471	163	4	.	.	PUNCT
easat-9471	164	1	the	the	DET
easat-9471	164	2	epoch	epoch	NOUN
easat-9471	164	3	is	be	AUX
easat-9471	164	4	1.000	1.000	NUM
easat-9471	164	5	.	.	PUNCT
easat-9471	165	1	and	and	CCONJ
easat-9471	165	2	the	the	DET
easat-9471	165	3	specific	specific	ADJ
easat-9471	165	4	hyperparameter	hyperparameter	NOUN
easat-9471	165	5	settings	setting	NOUN
easat-9471	165	6	are	be	AUX
easat-9471	165	7	shown	show	VERB
easat-9471	165	8	in	in	ADP
easat-9471	165	9	table	table	NOUN
easat-9471	165	10	2	2	NUM
easat-9471	165	11	.	.	PUNCT
easat-9471	165	12	table	table	NOUN
easat-9471	165	13	2	2	NUM
easat-9471	165	14	.	.	X
easat-9471	165	15	collaborative	collaborative	ADJ
easat-9471	165	16	filtering	filtering	NOUN
easat-9471	165	17	models	model	NOUN
easat-9471	165	18	and	and	CCONJ
easat-9471	165	19	grpo	grpo	NOUN
easat-9471	165	20	learning	learn	VERB
easat-9471	165	21	hyperparameters	hyperparameter	NOUN
easat-9471	165	22	.	.	PUNCT
easat-9471	166	1	datasets	dataset	NOUN
easat-9471	166	2	learning	learn	VERB
easat-9471	166	3	rate	rate	NOUN
easat-9471	166	4	dropout	dropout	NOUN
easat-9471	166	5	embedding	embed	VERB
easat-9471	166	6	dimension	dimension	NOUN
easat-9471	166	7	𝒅	𝒅	DET
easat-9471	166	8	sample	sample	NOUN
easat-9471	166	9	candidates	candidate	NOUN
easat-9471	166	10	𝑮	𝑮	ADP
easat-9471	166	11	𝜷	𝜷	NOUN
easat-9471	166	12	𝜺	𝜺	NOUN
easat-9471	166	13	baby	baby	NOUN
easat-9471	166	14	products	product	NOUN
easat-9471	166	15	0.0001	0.0001	NUM
easat-9471	166	16	0.5	0.5	NUM
easat-9471	166	17	64	64	NUM
easat-9471	166	18	4096	4096	NUM
easat-9471	166	19	0.01	0.01	NUM
easat-9471	166	20	0.1	0.1	NUM
easat-9471	166	21	video	video	NOUN
easat-9471	166	22	games	game	NOUN
easat-9471	166	23	0.0001	0.0001	NUM
easat-9471	166	24	0.5	0.5	NUM
easat-9471	166	25	64	64	NUM
easat-9471	166	26	4096	4096	NUM
easat-9471	166	27	0.01	0.01	NUM
easat-9471	166	28	0.1	0.1	NUM
easat-9471	166	29	industrial	industrial	ADJ
easat-9471	166	30	&	&	CCONJ
easat-9471	166	31	scientific	scientific	ADJ
easat-9471	166	32	0.0001	0.0001	NUM
easat-9471	166	33	0.5	0.5	NUM
easat-9471	166	34	64	64	NUM
easat-9471	166	35	4096	4096	NUM
easat-9471	166	36	0.01	0.01	NUM
easat-9471	166	37	0.1	0.1	NUM
easat-9471	166	38	4.2	4.2	NUM
easat-9471	166	39	.	.	PUNCT
easat-9471	167	1	performance	performance	NOUN
easat-9471	167	2	comparison	comparison	NOUN
easat-9471	167	3	the	the	DET
easat-9471	167	4	experimental	experimental	ADJ
easat-9471	167	5	results	result	NOUN
easat-9471	167	6	show	show	VERB
easat-9471	167	7	that	that	SCONJ
easat-9471	167	8	optimizing	optimize	VERB
easat-9471	167	9	the	the	DET
easat-9471	167	10	collaborative	collaborative	ADJ
easat-9471	167	11	filtering	filtering	NOUN
easat-9471	167	12	model	model	NOUN
easat-9471	167	13	using	use	VERB
easat-9471	167	14	the	the	DET
easat-9471	167	15	grpo	grpo	NOUN
easat-9471	167	16	algorithm	algorithm	NOUN
easat-9471	167	17	yields	yield	VERB
easat-9471	167	18	different	different	ADJ
easat-9471	167	19	performance	performance	NOUN
easat-9471	167	20	depending	depend	VERB
easat-9471	167	21	on	on	ADP
easat-9471	167	22	the	the	DET
easat-9471	167	23	model	model	NOUN
easat-9471	167	24	architecture	architecture	NOUN
easat-9471	167	25	.	.	PUNCT
easat-9471	168	1	as	as	SCONJ
easat-9471	168	2	shown	show	VERB
easat-9471	168	3	in	in	ADP
easat-9471	168	4	table	table	NOUN
easat-9471	168	5	3	3	NUM
easat-9471	168	6	.	.	PUNCT
easat-9471	168	7	𝐺𝑆𝐴𝑈	𝐺𝑆𝐴𝑈	NOUN
easat-9471	168	8	−	−	NOUN
easat-9471	168	9	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑.	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑.	PRON
easat-9471	168	10	in	in	ADP
easat-9471	168	11	which	which	DET
easat-9471	168	12	grpo	grpo	NOUN
easat-9471	168	13	is	be	AUX
easat-9471	168	14	applied	apply	VERB
easat-9471	168	15	to	to	ADP
easat-9471	168	16	the	the	DET
easat-9471	168	17	integrated	integrate	VERB
easat-9471	168	18	sequential	sequential	ADJ
easat-9471	168	19	-	-	PUNCT
easat-9471	168	20	graph	graph	NOUN
easat-9471	168	21	model	model	NOUN
easat-9471	168	22	gsau	gsau	NOUN
easat-9471	168	23	.	.	PUNCT
easat-9471	169	1	consistently	consistently	ADV
easat-9471	169	2	improves	improve	VERB
easat-9471	169	3	recommendation	recommendation	NOUN
easat-9471	169	4	performance	performance	NOUN
easat-9471	169	5	across	across	ADP
easat-9471	169	6	all	all	DET
easat-9471	169	7	datasets	dataset	NOUN
easat-9471	169	8	.	.	PUNCT
easat-9471	170	1	on	on	ADP
easat-9471	170	2	the	the	DET
easat-9471	170	3	industrial	industrial	ADJ
easat-9471	170	4	and	and	CCONJ
easat-9471	170	5	scientific	scientific	ADJ
easat-9471	170	6	and	and	CCONJ
easat-9471	170	7	video	video	NOUN
easat-9471	170	8	games	game	NOUN
easat-9471	170	9	datasets	dataset	NOUN
easat-9471	170	10	.	.	PUNCT
easat-9471	171	1	𝐺𝑆𝐴𝑈	𝐺𝑆𝐴𝑈	NOUN
easat-9471	171	2	−	−	VERB
easat-9471	171	3	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑	NOUN
easat-9471	171	4	yields	yield	NOUN
easat-9471	171	5	modest	modest	ADJ
easat-9471	171	6	relative	relative	ADJ
easat-9471	171	7	gains	gain	NOUN
easat-9471	171	8	of	of	ADP
easat-9471	171	9	approximately	approximately	ADV
easat-9471	171	10	1.5%-3	1.5%-3	NUM
easat-9471	171	11	%	%	NOUN
easat-9471	171	12	.	.	PUNCT
easat-9471	172	1	most	most	ADV
easat-9471	172	2	notably	notably	ADV
easat-9471	172	3	.	.	PUNCT
easat-9471	173	1	on	on	ADP
easat-9471	173	2	the	the	DET
easat-9471	173	3	baby	baby	NOUN
easat-9471	173	4	products	product	NOUN
easat-9471	173	5	dataset	dataset	VERB
easat-9471	173	6	it	it	PRON
easat-9471	173	7	achieves	achieve	VERB
easat-9471	173	8	substantial	substantial	ADJ
easat-9471	173	9	improvement	improvement	NOUN
easat-9471	173	10	of	of	ADP
easat-9471	173	11	about	about	ADP
easat-9471	173	12	13%-15	13%-15	NUM
easat-9471	173	13	%	%	NOUN
easat-9471	173	14	.	.	PUNCT
easat-9471	174	1	with	with	ADP
easat-9471	174	2	the	the	DET
easat-9471	174	3	largest	large	ADJ
easat-9471	174	4	performance	performance	NOUN
easat-9471	174	5	improvement	improvement	NOUN
easat-9471	174	6	of	of	ADP
easat-9471	174	7	about	about	ADP
easat-9471	174	8	15.16	15.16	NUM
easat-9471	174	9	%	%	NOUN
easat-9471	174	10	in	in	ADP
easat-9471	174	11	the	the	DET
easat-9471	174	12	recall@10	recall@10	NOUN
easat-9471	174	13	.	.	PUNCT
easat-9471	175	1	likewise	likewise	ADV
easat-9471	175	2	.	.	PUNCT
easat-9471	176	1	𝐿𝑖𝑔ℎ𝑡𝐺𝐶𝑁	𝐿𝑖𝑔ℎ𝑡𝐺𝐶𝑁	VERB
easat-9471	177	1	−	−	PROPN
easat-9471	177	2	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑.	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑.	PRON
easat-9471	177	3	which	which	PRON
easat-9471	177	4	optimized	optimize	VERB
easat-9471	177	5	the	the	DET
easat-9471	177	6	graph	graph	NOUN
easat-9471	177	7	-	-	PUNCT
easat-9471	177	8	based	base	VERB
easat-9471	177	9	model	model	NOUN
easat-9471	177	10	lightgcn	lightgcn	NOUN
easat-9471	177	11	with	with	ADP
easat-9471	177	12	grpo	grpo	NOUN
easat-9471	177	13	.	.	PUNCT
easat-9471	177	14	showed	show	VERB
easat-9471	177	15	a	a	DET
easat-9471	177	16	slight	slight	ADJ
easat-9471	177	17	performance	performance	NOUN
easat-9471	177	18	improvement	improvement	NOUN
easat-9471	177	19	of	of	ADP
easat-9471	177	20	approximately	approximately	ADV
easat-9471	177	21	0.07%-0.81	0.07%-0.81	ADJ
easat-9471	177	22	%	%	NOUN
easat-9471	177	23	on	on	ADP
easat-9471	177	24	the	the	DET
easat-9471	177	25	industrial	industrial	ADJ
easat-9471	177	26	and	and	CCONJ
easat-9471	177	27	scientific	scientific	ADJ
easat-9471	177	28	and	and	CCONJ
easat-9471	177	29	video	video	NOUN
easat-9471	177	30	games	game	NOUN
easat-9471	177	31	datasets	dataset	NOUN
easat-9471	177	32	.	.	PUNCT
easat-9471	178	1	while	while	SCONJ
easat-9471	178	2	exhibiting	exhibit	VERB
easat-9471	178	3	no	no	DET
easat-9471	178	4	meaningful	meaningful	ADJ
easat-9471	178	5	change	change	NOUN
easat-9471	178	6	on	on	ADP
easat-9471	178	7	the	the	DET
easat-9471	178	8	baby	baby	NOUN
easat-9471	178	9	products	product	NOUN
easat-9471	178	10	dataset	dataset	VERB
easat-9471	178	11	.	.	PUNCT
easat-9471	179	1	in	in	ADP
easat-9471	179	2	contrast	contrast	NOUN
easat-9471	179	3	.	.	PUNCT
easat-9471	180	1	𝑆𝐴𝑆𝑅𝑒𝑐	𝑆𝐴𝑆𝑅𝑒𝑐	PROPN
easat-9471	180	2	−	−	PROPN
easat-9471	180	3	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑.	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑.	NOUN
easat-9471	180	4	in	in	ADP
easat-9471	180	5	which	which	DET
easat-9471	180	6	grpo	grpo	NOUN
easat-9471	180	7	is	be	AUX
easat-9471	180	8	applied	apply	VERB
easat-9471	180	9	to	to	ADP
easat-9471	180	10	the	the	DET
easat-9471	180	11	sequential	sequential	ADJ
easat-9471	180	12	model	model	NOUN
easat-9471	180	13	sasrec	sasrec	NOUN
easat-9471	180	14	.	.	PUNCT
easat-9471	181	1	suffers	suffer	VERB
easat-9471	181	2	performance	performance	NOUN
easat-9471	181	3	degradation	degradation	NOUN
easat-9471	181	4	on	on	ADP
easat-9471	181	5	every	every	DET
easat-9471	181	6	dataset	dataset	NOUN
easat-9471	181	7	.	.	PUNCT
easat-9471	182	1	in	in	ADP
easat-9471	182	2	particular	particular	ADJ
easat-9471	182	3	.	.	PUNCT
easat-9471	183	1	on	on	ADP
easat-9471	183	2	the	the	DET
easat-9471	183	3	video	video	NOUN
easat-9471	183	4	games	game	NOUN
easat-9471	183	5	dataset	dataset	VERB
easat-9471	183	6	.	.	PUNCT
easat-9471	184	1	ndcg@10	ndcg@10	PROPN
easat-9471	184	2	suffers	suffer	VERB
easat-9471	184	3	the	the	DET
easat-9471	184	4	most	most	ADV
easat-9471	184	5	significant	significant	ADJ
easat-9471	184	6	drop	drop	NOUN
easat-9471	184	7	.	.	PUNCT
easat-9471	185	1	decreasing	decrease	VERB
easat-9471	185	2	by	by	ADP
easat-9471	185	3	38.3	38.3	NUM
easat-9471	185	4	%	%	NOUN
easat-9471	185	5	.	.	PUNCT
easat-9471	186	1	this	this	PRON
easat-9471	186	2	suggests	suggest	VERB
easat-9471	186	3	that	that	SCONJ
easat-9471	186	4	the	the	DET
easat-9471	186	5	policy	policy	NOUN
easat-9471	186	6	gradient	gradient	NOUN
easat-9471	186	7	update	update	NOUN
easat-9471	186	8	process	process	NOUN
easat-9471	186	9	for	for	ADP
easat-9471	186	10	item	item	NOUN
easat-9471	186	11	and	and	CCONJ
easat-9471	186	12	positional	positional	ADJ
easat-9471	186	13	embeddings	embedding	NOUN
easat-9471	186	14	in	in	ADP
easat-9471	186	15	sequential	sequential	ADJ
easat-9471	186	16	collaborative	collaborative	ADJ
easat-9471	186	17	filtering	filtering	NOUN
easat-9471	186	18	models	model	NOUN
easat-9471	186	19	may	may	AUX
easat-9471	186	20	differ	differ	VERB
easat-9471	186	21	from	from	ADP
easat-9471	186	22	that	that	PRON
easat-9471	186	23	in	in	ADP
easat-9471	186	24	graph	graph	NOUN
easat-9471	186	25	and	and	CCONJ
easat-9471	186	26	integrated	integrated	ADJ
easat-9471	186	27	models	model	NOUN
easat-9471	186	28	.	.	PUNCT
easat-9471	187	1	878	878	NUM
easat-9471	187	2	edelweiss	edelweiss	PROPN
easat-9471	187	3	applied	apply	VERB
easat-9471	187	4	science	science	NOUN
easat-9471	187	5	and	and	CCONJ
easat-9471	187	6	technology	technology	NOUN
easat-9471	187	7	issn	issn	PROPN
easat-9471	187	8	:	:	PUNCT
easat-9471	187	9	2576	2576	NUM
easat-9471	187	10	-	-	SYM
easat-9471	187	11	8484	8484	NUM
easat-9471	187	12	vol	vol	NOUN
easat-9471	187	13	.	.	PROPN
easat-9471	188	1	9	9	NUM
easat-9471	188	2	,	,	PUNCT
easat-9471	188	3	no	no	INTJ
easat-9471	188	4	.	.	NOUN
easat-9471	188	5	8	8	NUM
easat-9471	188	6	:	:	PUNCT
easat-9471	188	7	871	871	NUM
easat-9471	188	8	-	-	SYM
easat-9471	188	9	881	881	NUM
easat-9471	188	10	,	,	PUNCT
easat-9471	188	11	2025	2025	NUM
easat-9471	188	12	doi	doi	NOUN
easat-9471	188	13	:	:	PUNCT
easat-9471	188	14	10.55214/2576	10.55214/2576	NUM
easat-9471	188	15	-	-	PUNCT
easat-9471	188	16	8484.v9i8.9471	8484.v9i8.9471	NOUN
easat-9471	189	1	©	©	PROPN
easat-9471	189	2	2025	2025	NUM
easat-9471	189	3	by	by	ADP
easat-9471	189	4	the	the	DET
easat-9471	189	5	authors	author	NOUN
easat-9471	189	6	;	;	PUNCT
easat-9471	189	7	licensee	licensee	PROPN
easat-9471	189	8	learning	learn	VERB
easat-9471	189	9	gate	gate	NOUN
easat-9471	189	10	table	table	NOUN
easat-9471	189	11	3	3	NUM
easat-9471	189	12	.	.	PUNCT
easat-9471	189	13	comparison	comparison	NOUN
easat-9471	189	14	of	of	ADP
easat-9471	189	15	top	top	ADJ
easat-9471	189	16	recommendation	recommendation	NOUN
easat-9471	189	17	performance	performance	NOUN
easat-9471	189	18	between	between	ADP
easat-9471	189	19	our	our	PRON
easat-9471	189	20	method	method	NOUN
easat-9471	189	21	models	model	NOUN
easat-9471	189	22	and	and	CCONJ
easat-9471	189	23	baseline	baseline	NOUN
easat-9471	189	24	models	model	NOUN
easat-9471	189	25	and	and	CCONJ
easat-9471	189	26	grpo	grpo	NOUN
easat-9471	189	27	algorithms	algorithm	NOUN
easat-9471	189	28	applied	apply	VERB
easat-9471	189	29	to	to	ADP
easat-9471	189	30	amazon	amazon	NOUN
easat-9471	189	31	industrial	industrial	ADJ
easat-9471	189	32	and	and	CCONJ
easat-9471	189	33	scientific	scientific	ADJ
easat-9471	189	34	.	.	PUNCT
easat-9471	190	1	video	video	NOUN
easat-9471	190	2	games	game	NOUN
easat-9471	190	3	.	.	PUNCT
easat-9471	191	1	and	and	CCONJ
easat-9471	191	2	baby	baby	NOUN
easat-9471	191	3	products	product	NOUN
easat-9471	191	4	review	review	VERB
easat-9471	191	5	datasets	dataset	NOUN
easat-9471	191	6	.	.	PUNCT
easat-9471	192	1	r	r	NOUN
easat-9471	192	2	stands	stand	VERB
easat-9471	192	3	for	for	ADP
easat-9471	192	4	recall	recall	NOUN
easat-9471	192	5	.	.	PUNCT
easat-9471	193	1	and	and	CCONJ
easat-9471	193	2	n	n	PRON
easat-9471	193	3	stands	stand	VERB
easat-9471	193	4	for	for	ADP
easat-9471	193	5	ndcg	ndcg	NOUN
easat-9471	193	6	evaluation	evaluation	NOUN
easat-9471	193	7	metrics	metric	NOUN
easat-9471	193	8	.	.	PUNCT
easat-9471	194	1	superior	superior	ADJ
easat-9471	194	2	performance	performance	NOUN
easat-9471	194	3	is	be	AUX
easat-9471	194	4	indicated	indicate	VERB
easat-9471	194	5	in	in	ADP
easat-9471	194	6	bold	bold	ADJ
easat-9471	194	7	.	.	PUNCT
easat-9471	195	1	method	method	PROPN
easat-9471	195	2	baby	baby	NOUN
easat-9471	195	3	products	product	NOUN
easat-9471	195	4	video	video	NOUN
easat-9471	195	5	games	games	PROPN
easat-9471	195	6	industrial	industrial	ADJ
easat-9471	195	7	&	&	CCONJ
easat-9471	195	8	scientific	scientific	ADJ
easat-9471	195	9	r@10	r@10	NOUN
easat-9471	195	10	r@50	r@50	NUM
easat-9471	195	11	n@10	n@10	NUM
easat-9471	195	12	n@50	n@50	NUM
easat-9471	195	13	r@10	r@10	NOUN
easat-9471	195	14	r@50	r@50	NUM
easat-9471	195	15	n@10	n@10	NUM
easat-9471	195	16	n@50	n@50	NUM
easat-9471	195	17	r@10	r@10	NOUN
easat-9471	195	18	r@50	r@50	NUM
easat-9471	195	19	n@10	n@10	NUM
easat-9471	195	20	n@50	n@50	PRON
easat-9471	195	21	sasrec	sasrec	NOUN
easat-9471	195	22	0.0394	0.0394	NUM
easat-9471	195	23	0.0976	0.0976	NUM
easat-9471	195	24	0.0219	0.0219	NUM
easat-9471	195	25	0.0344	0.0344	NUM
easat-9471	195	26	0.1005	0.1005	NUM
easat-9471	195	27	0.2133	0.2133	NUM
easat-9471	195	28	0.0569	0.0569	NUM
easat-9471	195	29	0.0815	0.0815	NUM
easat-9471	195	30	0.0408	0.0408	NUM
easat-9471	195	31	0.0940	0.0940	NUM
easat-9471	195	32	0.0232	0.0232	NUM
easat-9471	195	33	0.0346	0.0346	NUM
easat-9471	195	34	𝑆𝐴𝑆𝑅𝑒𝑐	𝑆𝐴𝑆𝑅𝑒𝑐	PROPN
easat-9471	195	35	−	−	ADP
easat-9471	195	36	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑	NOUN
easat-9471	195	37	0.0306	0.0306	NUM
easat-9471	195	38	0.0858	0.0858	NUM
easat-9471	195	39	0.0164	0.0164	NUM
easat-9471	195	40	0.0282	0.0282	NUM
easat-9471	195	41	0.0639	0.0639	NUM
easat-9471	195	42	0.1529	0.1529	NUM
easat-9471	195	43	0.0351	0.0351	NUM
easat-9471	195	44	0.0543	0.0543	NUM
easat-9471	195	45	0.0389	0.0389	NUM
easat-9471	195	46	0.0976	0.0976	NUM
easat-9471	195	47	0.0202	0.0202	NUM
easat-9471	195	48	0.0329	0.0329	NUM
easat-9471	195	49	lightgcn	lightgcn	NOUN
easat-9471	195	50	0.0209	0.0209	NUM
easat-9471	195	51	0.0487	0.0487	NUM
easat-9471	195	52	0.0122	0.0122	NUM
easat-9471	195	53	0.0182	0.0182	NUM
easat-9471	195	54	0.0648	0.0648	NUM
easat-9471	195	55	0.1374	0.1374	NUM
easat-9471	195	56	0.0372	0.0372	NUM
easat-9471	195	57	0.0530	0.0530	NUM
easat-9471	195	58	0.0296	0.0296	NUM
easat-9471	195	59	0.0658	0.0658	NUM
easat-9471	195	60	0.0170	0.0170	NUM
easat-9471	195	61	0.0248	0.0248	NUM
easat-9471	195	62	𝐿𝑖𝑔ℎ𝑡𝐺𝐶𝑁	𝐿𝑖𝑔ℎ𝑡𝐺𝐶𝑁	VERB
easat-9471	195	63	−	−	PROPN
easat-9471	196	1	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑	SYM
easat-9471	196	2	0.0209	0.0209	NUM
easat-9471	196	3	0.0487	0.0487	NUM
easat-9471	196	4	0.0122	0.0122	NUM
easat-9471	196	5	0.0182	0.0182	NUM
easat-9471	196	6	0.0649	0.0649	NUM
easat-9471	196	7	0.1375	0.1375	NUM
easat-9471	196	8	0.0372	0.0372	NUM
easat-9471	196	9	0.0530	0.0530	NUM
easat-9471	196	10	0.0297	0.0297	NUM
easat-9471	196	11	0.0660	0.0660	NUM
easat-9471	196	12	0.0171	0.0171	NUM
easat-9471	196	13	0.0250	0.0250	NUM
easat-9471	196	14	gsau	gsau	ADJ
easat-9471	196	15	0.0322	0.0322	NUM
easat-9471	196	16	0.0805	0.0805	NUM
easat-9471	197	1	0.0177	0.0177	NUM
easat-9471	197	2	0.0281	0.0281	NUM
easat-9471	197	3	0.0967	0.0967	NUM
easat-9471	197	4	0.2139	0.2139	NUM
easat-9471	197	5	0.0546	0.0546	NUM
easat-9471	197	6	0.0801	0.0801	NUM
easat-9471	197	7	0.0459	0.0459	NUM
easat-9471	197	8	0.1037	0.1037	NUM
easat-9471	197	9	0.0263	0.0263	NUM
easat-9471	197	10	0.0387	0.0387	NUM
easat-9471	197	11	𝐺𝑆𝐴𝑈	𝐺𝑆𝐴𝑈	NOUN
easat-9471	197	12	−	−	NOUN
easat-9471	197	13	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑	𝐺𝑅𝑃𝑂𝑜𝑢𝑟𝑚𝑒𝑡ℎ𝑜𝑑	SYM
easat-9471	197	14	0.0366	0.0366	NUM
easat-9471	197	15	0.0927	0.0927	NUM
easat-9471	197	16	0.0200	0.0200	NUM
easat-9471	197	17	0.0321	0.0321	NUM
easat-9471	197	18	0.0995	0.0995	NUM
easat-9471	197	19	0.2184	0.2184	NUM
easat-9471	197	20	0.0555	0.0555	NUM
easat-9471	197	21	0.0813	0.0813	NUM
easat-9471	197	22	0.0468	0.0468	NUM
easat-9471	197	23	0.1063	0.1063	NUM
easat-9471	197	24	0.0267	0.0267	NUM
easat-9471	197	25	0.0395	0.0395	NUM
easat-9471	197	26	table	table	NOUN
easat-9471	197	27	4	4	NUM
easat-9471	197	28	.	.	PUNCT
easat-9471	197	29	performance	performance	NOUN
easat-9471	197	30	differences	difference	NOUN
easat-9471	197	31	between	between	ADP
easat-9471	197	32	policy	policy	NOUN
easat-9471	197	33	gradient	gradient	NOUN
easat-9471	197	34	update	update	NOUN
easat-9471	197	35	targets	target	NOUN
easat-9471	197	36	.	.	PUNCT
easat-9471	198	1	bold	bold	ADJ
easat-9471	198	2	text	text	NOUN
easat-9471	198	3	indicates	indicate	VERB
easat-9471	198	4	the	the	DET
easat-9471	198	5	highest	high	ADJ
easat-9471	198	6	performance	performance	NOUN
easat-9471	198	7	.	.	PUNCT
easat-9471	199	1	and	and	CCONJ
easat-9471	199	2	underlined	underline	VERB
easat-9471	199	3	text	text	NOUN
easat-9471	199	4	indicates	indicate	VERB
easat-9471	199	5	the	the	DET
easat-9471	199	6	second	second	ADJ
easat-9471	199	7	highest	high	ADJ
easat-9471	199	8	performance	performance	NOUN
easat-9471	199	9	.	.	PUNCT
easat-9471	200	1	method	method	NOUN
easat-9471	200	2	baby	baby	NOUN
easat-9471	200	3	products	product	NOUN
easat-9471	200	4	video	video	NOUN
easat-9471	200	5	games	games	PROPN
easat-9471	200	6	industrial	industrial	ADJ
easat-9471	200	7	&	&	CCONJ
easat-9471	200	8	scientific	scientific	ADJ
easat-9471	200	9	r@10	r@10	NOUN
easat-9471	200	10	r@50	r@50	NUM
easat-9471	200	11	n@10	n@10	NUM
easat-9471	200	12	n@50	n@50	NUM
easat-9471	200	13	r@10	r@10	NOUN
easat-9471	200	14	r@50	r@50	NUM
easat-9471	200	15	n@10	n@10	NUM
easat-9471	200	16	n@50	n@50	NUM
easat-9471	200	17	r@10	r@10	NOUN
easat-9471	200	18	r@50	r@50	NUM
easat-9471	200	19	n@10	n@10	PUNCT
easat-9471	200	20	n@50	n@50	X
easat-9471	200	21	gsau	gsau	PROPN
easat-9471	200	22	0.0322	0.0322	NUM
easat-9471	200	23	0.0805	0.0805	NUM
easat-9471	201	1	0.0177	0.0177	NUM
easat-9471	201	2	0.0281	0.0281	NUM
easat-9471	201	3	0.967	0.967	NUM
easat-9471	201	4	0.2139	0.2139	NUM
easat-9471	201	5	0.0546	0.0546	NUM
easat-9471	201	6	0.0801	0.0801	NUM
easat-9471	201	7	0.0459	0.0459	NUM
easat-9471	201	8	0.1037	0.1037	NUM
easat-9471	201	9	0.0263	0.0263	NUM
easat-9471	201	10	0.0387	0.0387	NUM
easat-9471	201	11	𝑒𝑖𝑡𝑒𝑚	𝑒𝑖𝑡𝑒𝑚	NOUN
easat-9471	201	12	0.0364	0.0364	NUM
easat-9471	202	1	0.0921	0.0921	NUM
easat-9471	202	2	0.0199	0.0199	NUM
easat-9471	202	3	0.0319	0.0319	NUM
easat-9471	202	4	0.0993	0.0993	NUM
easat-9471	202	5	0.2181	0.2181	NUM
easat-9471	202	6	0.0554	0.0554	NUM
easat-9471	202	7	0.0812	0.0812	NUM
easat-9471	202	8	0.0467	0.0467	NUM
easat-9471	202	9	0.1062	0.1062	NUM
easat-9471	202	10	0.0267	0.0267	NUM
easat-9471	202	11	0.0395	0.0395	NUM
easat-9471	202	12	𝑒𝑖𝑡𝑒𝑚	𝑒𝑖𝑡𝑒𝑚	ADV
easat-9471	202	13	.	.	PUNCT
easat-9471	203	1	𝑒𝑢𝑠𝑒𝑟	𝑒𝑢𝑠𝑒𝑟	PROPN
easat-9471	203	2	0.0366	0.0366	NUM
easat-9471	204	1	0.0927	0.0927	NUM
easat-9471	204	2	0.0200	0.0200	NUM
easat-9471	204	3	0.0321	0.0321	NUM
easat-9471	204	4	0.0995	0.0995	NUM
easat-9471	204	5	0.2184	0.2184	NUM
easat-9471	204	6	0.0555	0.0555	NUM
easat-9471	204	7	0.0813	0.0813	NUM
easat-9471	204	8	0.0468	0.0468	NUM
easat-9471	204	9	0.1063	0.1063	NUM
easat-9471	204	10	0.0267	0.0267	NUM
easat-9471	204	11	0.0395	0.0395	NUM
easat-9471	204	12	𝑒𝑖𝑡𝑒𝑚	𝑒𝑖𝑡𝑒𝑚	ADV
easat-9471	204	13	.	.	PUNCT
easat-9471	205	1	𝑒𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛	𝑒𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛	VERB
easat-9471	205	2	0.0314	0.0314	NUM
easat-9471	205	3	0.0776	0.0776	NUM
easat-9471	205	4	0.0173	0.0173	NUM
easat-9471	205	5	0.0272	0.0272	NUM
easat-9471	205	6	0.0875	0.0875	NUM
easat-9471	205	7	0.1969	0.1969	NUM
easat-9471	205	8	0.0486	0.0486	NUM
easat-9471	205	9	0.0723	0.0723	NUM
easat-9471	205	10	0.0451	0.0451	NUM
easat-9471	205	11	0.1019	0.1019	NUM
easat-9471	205	12	0.0257	0.0257	NUM
easat-9471	205	13	0.0380	0.0380	NUM
easat-9471	205	14	𝑒𝑖𝑡𝑒𝑚	𝑒𝑖𝑡𝑒𝑚	NOUN
easat-9471	205	15	.	.	PUNCT
easat-9471	206	1	𝑒𝑢𝑠𝑒𝑟	𝑒𝑢𝑠𝑒𝑟	PROPN
easat-9471	206	2	.	.	PUNCT
easat-9471	207	1	𝑒𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛	𝑒𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛	NOUN
easat-9471	207	2	0.0317	0.0317	NUM
easat-9471	207	3	0.0776	0.0776	NUM
easat-9471	208	1	0.0174	0.0174	NUM
easat-9471	208	2	0.0273	0.0273	NUM
easat-9471	208	3	0.0876	0.0876	NUM
easat-9471	208	4	0.1972	0.1972	NUM
easat-9471	208	5	0.0487	0.0487	NUM
easat-9471	208	6	0.0725	0.0725	NUM
easat-9471	208	7	0.0451	0.0451	NUM
easat-9471	208	8	0.1021	0.1021	NUM
easat-9471	208	9	0.0258	0.0258	NUM
easat-9471	208	10	0.0381	0.0381	NUM
easat-9471	208	11	879	879	NUM
easat-9471	208	12	edelweiss	edelweiss	PROPN
easat-9471	208	13	applied	apply	VERB
easat-9471	208	14	science	science	NOUN
easat-9471	208	15	and	and	CCONJ
easat-9471	208	16	technology	technology	NOUN
easat-9471	208	17	issn	issn	PROPN
easat-9471	208	18	:	:	PUNCT
easat-9471	208	19	2576	2576	NUM
easat-9471	208	20	-	-	SYM
easat-9471	208	21	8484	8484	NUM
easat-9471	208	22	vol	vol	NOUN
easat-9471	208	23	.	.	PROPN
easat-9471	209	1	9	9	NUM
easat-9471	209	2	,	,	PUNCT
easat-9471	209	3	no	no	INTJ
easat-9471	209	4	.	.	NOUN
easat-9471	209	5	8	8	NUM
easat-9471	209	6	:	:	PUNCT
easat-9471	209	7	871	871	NUM
easat-9471	209	8	-	-	SYM
easat-9471	209	9	881	881	NUM
easat-9471	209	10	,	,	PUNCT
easat-9471	209	11	2025	2025	NUM
easat-9471	209	12	doi	doi	NOUN
easat-9471	209	13	:	:	PUNCT
easat-9471	209	14	10.55214/2576	10.55214/2576	NUM
easat-9471	209	15	-	-	PUNCT
easat-9471	209	16	8484.v9i8.9471	8484.v9i8.9471	NOUN
easat-9471	210	1	©	©	PROPN
easat-9471	210	2	2025	2025	NUM
easat-9471	210	3	by	by	ADP
easat-9471	210	4	the	the	DET
easat-9471	210	5	authors	author	NOUN
easat-9471	210	6	;	;	PUNCT
easat-9471	210	7	licensee	licensee	PROPN
easat-9471	210	8	learning	learning	NOUN
easat-9471	210	9	gate	gate	NOUN
easat-9471	210	10	4.3	4.3	NUM
easat-9471	210	11	.	.	PUNCT
easat-9471	210	12	ablation	ablation	NOUN
easat-9471	210	13	study	study	NOUN
easat-9471	210	14	in	in	ADP
easat-9471	210	15	this	this	DET
easat-9471	210	16	section	section	NOUN
easat-9471	210	17	.	.	PUNCT
easat-9471	211	1	we	we	PRON
easat-9471	211	2	conducted	conduct	VERB
easat-9471	211	3	the	the	DET
easat-9471	211	4	ablation	ablation	NOUN
easat-9471	211	5	study	study	NOUN
easat-9471	211	6	to	to	PART
easat-9471	211	7	verify	verify	VERB
easat-9471	211	8	the	the	DET
easat-9471	211	9	performance	performance	NOUN
easat-9471	211	10	changes	change	NOUN
easat-9471	211	11	according	accord	VERB
easat-9471	211	12	to	to	ADP
easat-9471	211	13	the	the	DET
easat-9471	211	14	collaborative	collaborative	ADJ
easat-9471	211	15	filtering	filtering	NOUN
easat-9471	211	16	model	model	NOUN
easat-9471	211	17	architecture	architecture	NOUN
easat-9471	211	18	when	when	SCONJ
easat-9471	211	19	updating	update	VERB
easat-9471	211	20	the	the	DET
easat-9471	211	21	grpo	grpo	NOUN
easat-9471	211	22	policy	policy	NOUN
easat-9471	211	23	.	.	PUNCT
easat-9471	212	1	we	we	PRON
easat-9471	212	2	selectively	selectively	ADV
easat-9471	212	3	used	use	VERB
easat-9471	212	4	item	item	NOUN
easat-9471	212	5	embeddings	embedding	NOUN
easat-9471	212	6	.	.	PUNCT
easat-9471	213	1	user	user	NOUN
easat-9471	213	2	embeddings	embedding	NOUN
easat-9471	213	3	.	.	PUNCT
easat-9471	214	1	and	and	CCONJ
easat-9471	214	2	positional	positional	ADJ
easat-9471	214	3	embeddings	embedding	NOUN
easat-9471	214	4	.	.	PUNCT
easat-9471	215	1	which	which	PRON
easat-9471	215	2	constitute	constitute	VERB
easat-9471	215	3	the	the	DET
easat-9471	215	4	integrated	integrate	VERB
easat-9471	215	5	collaborative	collaborative	ADJ
easat-9471	215	6	filtering	filtering	NOUN
easat-9471	215	7	model	model	NOUN
easat-9471	215	8	gsau	gsau	NOUN
easat-9471	215	9	.	.	PROPN
easat-9471	215	10	for	for	ADP
easat-9471	215	11	policy	policy	NOUN
easat-9471	215	12	optimization	optimization	NOUN
easat-9471	215	13	.	.	PUNCT
easat-9471	216	1	as	as	SCONJ
easat-9471	216	2	shown	show	VERB
easat-9471	216	3	in	in	ADP
easat-9471	216	4	table	table	NOUN
easat-9471	216	5	4	4	NUM
easat-9471	216	6	.	.	PUNCT
easat-9471	217	1	when	when	SCONJ
easat-9471	217	2	only	only	ADJ
easat-9471	217	3	item	item	NOUN
easat-9471	217	4	embeddings	embedding	NOUN
easat-9471	217	5	were	be	AUX
easat-9471	217	6	used	use	VERB
easat-9471	217	7	for	for	ADP
easat-9471	217	8	policy	policy	NOUN
easat-9471	217	9	gradient	gradient	NOUN
easat-9471	217	10	updates	update	VERB
easat-9471	217	11	.	.	PUNCT
easat-9471	218	1	and	and	CCONJ
easat-9471	218	2	when	when	SCONJ
easat-9471	218	3	both	both	DET
easat-9471	218	4	item	item	NOUN
easat-9471	218	5	embeddings	embedding	NOUN
easat-9471	218	6	and	and	CCONJ
easat-9471	218	7	user	user	NOUN
easat-9471	218	8	embeddings	embedding	NOUN
easat-9471	218	9	were	be	AUX
easat-9471	218	10	used	use	VERB
easat-9471	218	11	for	for	ADP
easat-9471	218	12	policy	policy	NOUN
easat-9471	218	13	gradient	gradient	NOUN
easat-9471	218	14	updates	update	VERB
easat-9471	218	15	.	.	PUNCT
easat-9471	219	1	consistent	consistent	ADJ
easat-9471	219	2	performance	performance	NOUN
easat-9471	219	3	improvements	improvement	NOUN
easat-9471	219	4	were	be	AUX
easat-9471	219	5	observed	observe	VERB
easat-9471	219	6	across	across	ADP
easat-9471	219	7	all	all	DET
easat-9471	219	8	datasets	dataset	NOUN
easat-9471	219	9	—	—	PUNCT
easat-9471	219	10	baby	baby	NOUN
easat-9471	219	11	products	product	NOUN
easat-9471	219	12	.	.	PUNCT
easat-9471	220	1	video	video	NOUN
easat-9471	220	2	games	game	NOUN
easat-9471	220	3	.	.	PUNCT
easat-9471	221	1	and	and	CCONJ
easat-9471	221	2	industrial	industrial	ADJ
easat-9471	221	3	and	and	CCONJ
easat-9471	221	4	scientific	scientific	ADJ
easat-9471	221	5	—	—	PUNCT
easat-9471	221	6	compared	compare	VERB
easat-9471	221	7	to	to	ADP
easat-9471	221	8	the	the	DET
easat-9471	221	9	baseline	baseline	PROPN
easat-9471	221	10	model	model	NOUN
easat-9471	221	11	gsau	gsau	NOUN
easat-9471	221	12	.	.	PUNCT
easat-9471	222	1	in	in	ADP
easat-9471	222	2	particular	particular	ADJ
easat-9471	222	3	.	.	PUNCT
easat-9471	223	1	the	the	DET
easat-9471	223	2	greatest	great	ADJ
easat-9471	223	3	performance	performance	NOUN
easat-9471	223	4	improvement	improvement	NOUN
easat-9471	223	5	was	be	AUX
easat-9471	223	6	observed	observe	VERB
easat-9471	223	7	when	when	SCONJ
easat-9471	223	8	both	both	DET
easat-9471	223	9	item	item	NOUN
easat-9471	223	10	embeddings	embedding	NOUN
easat-9471	223	11	and	and	CCONJ
easat-9471	223	12	user	user	NOUN
easat-9471	223	13	embeddings	embedding	NOUN
easat-9471	223	14	were	be	AUX
easat-9471	223	15	used	use	VERB
easat-9471	223	16	for	for	ADP
easat-9471	223	17	policy	policy	NOUN
easat-9471	223	18	gradient	gradient	NOUN
easat-9471	223	19	updates	update	VERB
easat-9471	223	20	.	.	PUNCT
easat-9471	224	1	suggesting	suggest	VERB
easat-9471	224	2	that	that	SCONJ
easat-9471	224	3	user	user	NOUN
easat-9471	224	4	embeddings	embedding	NOUN
easat-9471	224	5	are	be	AUX
easat-9471	224	6	the	the	DET
easat-9471	224	7	most	most	ADV
easat-9471	224	8	significant	significant	ADJ
easat-9471	224	9	positive	positive	ADJ
easat-9471	224	10	factor	factor	NOUN
easat-9471	224	11	in	in	ADP
easat-9471	224	12	grpo	grpo	NOUN
easat-9471	224	13	reinforcement	reinforcement	NOUN
easat-9471	224	14	learning	learning	NOUN
easat-9471	224	15	.	.	PUNCT
easat-9471	225	1	in	in	ADP
easat-9471	225	2	contrast	contrast	NOUN
easat-9471	225	3	.	.	PUNCT
easat-9471	226	1	when	when	SCONJ
easat-9471	226	2	item	item	NOUN
easat-9471	226	3	embeddings	embedding	NOUN
easat-9471	226	4	.	.	PUNCT
easat-9471	227	1	user	user	NOUN
easat-9471	227	2	embeddings	embedding	NOUN
easat-9471	227	3	.	.	PUNCT
easat-9471	228	1	and	and	CCONJ
easat-9471	228	2	positional	positional	ADJ
easat-9471	228	3	embeddings	embedding	NOUN
easat-9471	228	4	were	be	AUX
easat-9471	228	5	all	all	PRON
easat-9471	228	6	used	use	VERB
easat-9471	228	7	in	in	ADP
easat-9471	228	8	policy	policy	NOUN
easat-9471	228	9	gradient	gradient	NOUN
easat-9471	228	10	updates	update	VERB
easat-9471	228	11	.	.	PUNCT
easat-9471	229	1	and	and	CCONJ
easat-9471	229	2	when	when	SCONJ
easat-9471	229	3	item	item	NOUN
easat-9471	229	4	embeddings	embedding	NOUN
easat-9471	229	5	and	and	CCONJ
easat-9471	229	6	positional	positional	ADJ
easat-9471	229	7	embeddings	embedding	NOUN
easat-9471	229	8	were	be	AUX
easat-9471	229	9	used	use	VERB
easat-9471	229	10	in	in	ADP
easat-9471	229	11	policy	policy	NOUN
easat-9471	229	12	gradient	gradient	NOUN
easat-9471	229	13	updates	update	VERB
easat-9471	229	14	.	.	PUNCT
easat-9471	230	1	consistent	consistent	ADJ
easat-9471	230	2	performance	performance	NOUN
easat-9471	230	3	degradation	degradation	NOUN
easat-9471	230	4	was	be	AUX
easat-9471	230	5	observed	observe	VERB
easat-9471	230	6	across	across	ADP
easat-9471	230	7	all	all	DET
easat-9471	230	8	datasets	dataset	NOUN
easat-9471	230	9	compared	compare	VERB
easat-9471	230	10	to	to	ADP
easat-9471	230	11	the	the	DET
easat-9471	230	12	baseline	baseline	PROPN
easat-9471	230	13	model	model	NOUN
easat-9471	230	14	gsau	gsau	NOUN
easat-9471	230	15	.	.	PUNCT
easat-9471	231	1	in	in	ADP
easat-9471	231	2	particular	particular	ADJ
easat-9471	231	3	.	.	PUNCT
easat-9471	232	1	when	when	SCONJ
easat-9471	232	2	item	item	NOUN
easat-9471	232	3	embeddings	embedding	NOUN
easat-9471	232	4	and	and	CCONJ
easat-9471	232	5	positional	positional	ADJ
easat-9471	232	6	embeddings	embedding	NOUN
easat-9471	232	7	were	be	AUX
easat-9471	232	8	used	use	VERB
easat-9471	232	9	together	together	ADV
easat-9471	232	10	for	for	SCONJ
easat-9471	232	11	policy	policy	NOUN
easat-9471	232	12	gradient	gradient	NOUN
easat-9471	232	13	updates	update	VERB
easat-9471	232	14	.	.	PUNCT
easat-9471	233	1	the	the	DET
easat-9471	233	2	largest	large	ADJ
easat-9471	233	3	performance	performance	NOUN
easat-9471	233	4	degradation	degradation	NOUN
easat-9471	233	5	was	be	AUX
easat-9471	233	6	observed	observe	VERB
easat-9471	233	7	.	.	PUNCT
easat-9471	234	1	suggesting	suggest	VERB
easat-9471	234	2	that	that	SCONJ
easat-9471	234	3	positional	positional	ADJ
easat-9471	234	4	embeddings	embedding	NOUN
easat-9471	234	5	are	be	AUX
easat-9471	234	6	the	the	DET
easat-9471	234	7	factor	factor	NOUN
easat-9471	234	8	that	that	PRON
easat-9471	234	9	has	have	VERB
easat-9471	234	10	the	the	DET
easat-9471	234	11	greatest	great	ADJ
easat-9471	234	12	negative	negative	ADJ
easat-9471	234	13	effect	effect	NOUN
easat-9471	234	14	in	in	ADP
easat-9471	234	15	grpo	grpo	NOUN
easat-9471	234	16	reinforcement	reinforcement	NOUN
easat-9471	234	17	learning	learning	NOUN
easat-9471	234	18	.	.	PUNCT
easat-9471	235	1	this	this	PRON
easat-9471	235	2	suggests	suggest	VERB
easat-9471	235	3	that	that	SCONJ
easat-9471	235	4	policy	policy	NOUN
easat-9471	235	5	gradient	gradient	NOUN
easat-9471	235	6	updates	update	NOUN
easat-9471	235	7	of	of	ADP
easat-9471	235	8	positional	positional	ADJ
easat-9471	235	9	embeddings	embedding	NOUN
easat-9471	235	10	may	may	AUX
easat-9471	235	11	destabilize	destabilize	VERB
easat-9471	235	12	the	the	DET
easat-9471	235	13	advantage	advantage	NOUN
easat-9471	235	14	estimates	estimate	NOUN
easat-9471	235	15	in	in	ADP
easat-9471	235	16	the	the	DET
easat-9471	235	17	grpo	grpo	NOUN
easat-9471	235	18	calculation	calculation	NOUN
easat-9471	235	19	process	process	NOUN
easat-9471	235	20	.	.	PUNCT
easat-9471	236	1	distorting	distort	VERB
easat-9471	236	2	the	the	DET
easat-9471	236	3	importance	importance	NOUN
easat-9471	236	4	ratio	ratio	NOUN
easat-9471	236	5	distribution	distribution	NOUN
easat-9471	236	6	.	.	PUNCT
easat-9471	237	1	these	these	DET
easat-9471	237	2	observations	observation	NOUN
easat-9471	237	3	demonstrate	demonstrate	VERB
easat-9471	237	4	that	that	SCONJ
easat-9471	237	5	embedding	embed	VERB
easat-9471	237	6	selection	selection	NOUN
easat-9471	237	7	is	be	AUX
easat-9471	237	8	as	as	ADV
easat-9471	237	9	important	important	ADJ
easat-9471	237	10	as	as	ADP
easat-9471	237	11	model	model	NOUN
easat-9471	237	12	architecture	architecture	NOUN
easat-9471	237	13	in	in	ADP
easat-9471	237	14	optimizing	optimize	VERB
easat-9471	237	15	grpo	grpo	NOUN
easat-9471	237	16	-	-	PUNCT
easat-9471	237	17	based	base	VERB
easat-9471	237	18	collaborative	collaborative	ADJ
easat-9471	237	19	filtering	filtering	NOUN
easat-9471	237	20	models	model	NOUN
easat-9471	237	21	.	.	PUNCT
easat-9471	238	1	in	in	ADP
easat-9471	238	2	particular	particular	ADJ
easat-9471	238	3	.	.	PUNCT
easat-9471	239	1	the	the	DET
easat-9471	239	2	significant	significant	ADJ
easat-9471	239	3	performance	performance	NOUN
easat-9471	239	4	degradation	degradation	NOUN
easat-9471	239	5	observed	observe	VERB
easat-9471	239	6	in	in	ADP
easat-9471	239	7	sequential	sequential	ADJ
easat-9471	239	8	models	model	NOUN
easat-9471	239	9	suggests	suggest	VERB
easat-9471	239	10	that	that	SCONJ
easat-9471	239	11	updated	update	VERB
easat-9471	239	12	positional	positional	ADJ
easat-9471	239	13	embeddings	embedding	NOUN
easat-9471	239	14	were	be	AUX
easat-9471	239	15	a	a	DET
easat-9471	239	16	major	major	ADJ
easat-9471	239	17	contributing	contribute	VERB
easat-9471	239	18	factor	factor	NOUN
easat-9471	239	19	.	.	PUNCT
easat-9471	240	1	5	5	X
easat-9471	240	2	.	.	X
easat-9471	240	3	conclusion	conclusion	NOUN
easat-9471	240	4	this	this	DET
easat-9471	240	5	paper	paper	NOUN
easat-9471	240	6	proposes	propose	VERB
easat-9471	240	7	a	a	DET
easat-9471	240	8	new	new	ADJ
easat-9471	240	9	method	method	NOUN
easat-9471	240	10	for	for	ADP
easat-9471	240	11	integrating	integrate	VERB
easat-9471	240	12	the	the	DET
easat-9471	240	13	grpo	grpo	NOUN
easat-9471	240	14	algorithm	algorithm	NOUN
easat-9471	240	15	into	into	ADP
easat-9471	240	16	a	a	DET
easat-9471	240	17	collaborative	collaborative	ADJ
easat-9471	240	18	filtering	filtering	NOUN
easat-9471	240	19	-	-	PUNCT
easat-9471	240	20	based	base	VERB
easat-9471	240	21	recommender	recommender	NOUN
easat-9471	240	22	system	system	NOUN
easat-9471	240	23	.	.	PUNCT
easat-9471	241	1	unlike	unlike	ADP
easat-9471	241	2	traditional	traditional	ADJ
easat-9471	241	3	collaborative	collaborative	ADJ
easat-9471	241	4	filtering	filtering	NOUN
easat-9471	241	5	models	model	NOUN
easat-9471	241	6	.	.	PUNCT
easat-9471	242	1	which	which	PRON
easat-9471	242	2	primarily	primarily	ADV
easat-9471	242	3	focus	focus	VERB
easat-9471	242	4	on	on	ADP
easat-9471	242	5	short	short	ADJ
easat-9471	242	6	-	-	PUNCT
easat-9471	242	7	term	term	NOUN
easat-9471	242	8	prediction	prediction	NOUN
easat-9471	242	9	accuracy	accuracy	NOUN
easat-9471	242	10	.	.	PUNCT
easat-9471	243	1	this	this	DET
easat-9471	243	2	paper	paper	NOUN
easat-9471	243	3	utilizes	utilize	VERB
easat-9471	243	4	reinforcement	reinforcement	NOUN
easat-9471	243	5	learning	learning	NOUN
easat-9471	243	6	to	to	PART
easat-9471	243	7	focus	focus	VERB
easat-9471	243	8	on	on	ADP
easat-9471	243	9	longterm	longterm	PROPN
easat-9471	243	10	user	user	NOUN
easat-9471	243	11	satisfaction	satisfaction	NOUN
easat-9471	243	12	and	and	CCONJ
easat-9471	243	13	prediction	prediction	NOUN
easat-9471	243	14	accuracy	accuracy	NOUN
easat-9471	243	15	.	.	PUNCT
easat-9471	244	1	achieving	achieve	VERB
easat-9471	244	2	performance	performance	NOUN
easat-9471	244	3	optimization	optimization	NOUN
easat-9471	244	4	using	use	VERB
easat-9471	244	5	various	various	ADJ
easat-9471	244	6	realworld	realworld	PROPN
easat-9471	244	7	datasets	dataset	NOUN
easat-9471	244	8	.	.	PUNCT
easat-9471	245	1	the	the	DET
easat-9471	245	2	grpo	grpo	NOUN
easat-9471	245	3	reinforcement	reinforcement	NOUN
easat-9471	245	4	learning	learning	NOUN
easat-9471	245	5	-	-	PUNCT
easat-9471	245	6	based	base	VERB
easat-9471	245	7	model	model	NOUN
easat-9471	245	8	achieved	achieve	VERB
easat-9471	245	9	superior	superior	ADJ
easat-9471	245	10	performance	performance	NOUN
easat-9471	245	11	compared	compare	VERB
easat-9471	245	12	to	to	ADP
easat-9471	245	13	the	the	DET
easat-9471	245	14	baseline	baseline	NOUN
easat-9471	245	15	model	model	NOUN
easat-9471	245	16	across	across	ADP
easat-9471	245	17	multiple	multiple	ADJ
easat-9471	245	18	metrics	metric	NOUN
easat-9471	245	19	.	.	PUNCT
easat-9471	246	1	notably	notably	ADV
easat-9471	246	2	.	.	PUNCT
easat-9471	247	1	it	it	PRON
easat-9471	247	2	demonstrated	demonstrate	VERB
easat-9471	247	3	up	up	ADP
easat-9471	247	4	to	to	ADP
easat-9471	247	5	a	a	DET
easat-9471	247	6	15	15	NUM
easat-9471	247	7	%	%	NOUN
easat-9471	247	8	performance	performance	NOUN
easat-9471	247	9	improvement	improvement	NOUN
easat-9471	247	10	on	on	ADP
easat-9471	247	11	the	the	DET
easat-9471	247	12	baby	baby	NOUN
easat-9471	247	13	products	product	NOUN
easat-9471	247	14	dataset	dataset	VERB
easat-9471	247	15	.	.	PUNCT
easat-9471	248	1	through	through	ADP
easat-9471	248	2	removal	removal	NOUN
easat-9471	248	3	experiments	experiment	NOUN
easat-9471	248	4	.	.	PUNCT
easat-9471	249	1	the	the	DET
easat-9471	249	2	role	role	NOUN
easat-9471	249	3	of	of	ADP
easat-9471	249	4	the	the	DET
easat-9471	249	5	grpo	grpo	NOUN
easat-9471	249	6	policy	policy	NOUN
easat-9471	249	7	gradient	gradient	NOUN
easat-9471	249	8	update	update	NOUN
easat-9471	249	9	target	target	NOUN
easat-9471	249	10	embeddings	embedding	NOUN
easat-9471	249	11	was	be	AUX
easat-9471	249	12	validated	validate	VERB
easat-9471	249	13	.	.	PUNCT
easat-9471	250	1	and	and	CCONJ
easat-9471	250	2	the	the	DET
easat-9471	250	3	most	most	ADV
easat-9471	250	4	outstanding	outstanding	ADJ
easat-9471	250	5	performance	performance	NOUN
easat-9471	250	6	optimization	optimization	NOUN
easat-9471	250	7	was	be	AUX
easat-9471	250	8	achieved	achieve	VERB
easat-9471	250	9	when	when	SCONJ
easat-9471	250	10	both	both	DET
easat-9471	250	11	item	item	NOUN
easat-9471	250	12	embeddings	embedding	NOUN
easat-9471	250	13	and	and	CCONJ
easat-9471	250	14	user	user	NOUN
easat-9471	250	15	embeddings	embedding	NOUN
easat-9471	250	16	were	be	AUX
easat-9471	250	17	updated	update	VERB
easat-9471	250	18	via	via	ADP
easat-9471	250	19	the	the	DET
easat-9471	250	20	policy	policy	NOUN
easat-9471	250	21	.	.	PUNCT
easat-9471	251	1	funding	funding	NOUN
easat-9471	251	2	:	:	PUNCT
easat-9471	251	3	this	this	DET
easat-9471	251	4	study	study	NOUN
easat-9471	251	5	was	be	AUX
easat-9471	251	6	financially	financially	ADV
easat-9471	251	7	supported	support	VERB
easat-9471	251	8	by	by	ADP
easat-9471	251	9	seoul	seoul	PROPN
easat-9471	251	10	national	national	PROPN
easat-9471	251	11	university	university	PROPN
easat-9471	251	12	of	of	ADP
easat-9471	251	13	science	science	NOUN
easat-9471	251	14	and	and	CCONJ
easat-9471	251	15	technology	technology	NOUN
easat-9471	251	16	.	.	PUNCT
easat-9471	252	1	transparency	transparency	NOUN
easat-9471	252	2	:	:	PUNCT
easat-9471	252	3	the	the	DET
easat-9471	252	4	authors	author	NOUN
easat-9471	252	5	confirm	confirm	VERB
easat-9471	252	6	that	that	SCONJ
easat-9471	252	7	the	the	DET
easat-9471	252	8	manuscript	manuscript	NOUN
easat-9471	252	9	is	be	AUX
easat-9471	252	10	an	an	DET
easat-9471	252	11	honest	hon	ADJ
easat-9471	252	12	.	.	PUNCT
easat-9471	253	1	accurate	accurate	ADJ
easat-9471	253	2	.	.	PUNCT
easat-9471	254	1	and	and	CCONJ
easat-9471	254	2	transparent	transparent	ADJ
easat-9471	254	3	account	account	NOUN
easat-9471	254	4	of	of	ADP
easat-9471	254	5	the	the	DET
easat-9471	254	6	study	study	NOUN
easat-9471	254	7	;	;	PUNCT
easat-9471	254	8	that	that	SCONJ
easat-9471	254	9	no	no	DET
easat-9471	254	10	vital	vital	ADJ
easat-9471	254	11	features	feature	NOUN
easat-9471	254	12	of	of	ADP
easat-9471	254	13	the	the	DET
easat-9471	254	14	study	study	NOUN
easat-9471	254	15	have	have	AUX
easat-9471	254	16	been	be	AUX
easat-9471	254	17	omitted	omit	VERB
easat-9471	254	18	;	;	PUNCT
easat-9471	254	19	and	and	CCONJ
easat-9471	254	20	that	that	SCONJ
easat-9471	254	21	any	any	DET
easat-9471	254	22	discrepancies	discrepancy	NOUN
easat-9471	254	23	from	from	ADP
easat-9471	254	24	the	the	DET
easat-9471	254	25	study	study	NOUN
easat-9471	254	26	as	as	SCONJ
easat-9471	254	27	planned	plan	VERB
easat-9471	254	28	have	have	AUX
easat-9471	254	29	been	be	AUX
easat-9471	254	30	explained	explain	VERB
easat-9471	254	31	.	.	PUNCT
easat-9471	255	1	this	this	DET
easat-9471	255	2	study	study	NOUN
easat-9471	255	3	followed	follow	VERB
easat-9471	255	4	all	all	DET
easat-9471	255	5	ethical	ethical	ADJ
easat-9471	255	6	practices	practice	NOUN
easat-9471	255	7	during	during	ADP
easat-9471	255	8	writing	writing	NOUN
easat-9471	255	9	.	.	PUNCT
easat-9471	256	1	copyright	copyright	NOUN
easat-9471	256	2	:	:	PUNCT
easat-9471	256	3	©	©	PROPN
easat-9471	256	4	2025	2025	NUM
easat-9471	256	5	by	by	ADP
easat-9471	256	6	the	the	DET
easat-9471	256	7	authors	author	NOUN
easat-9471	256	8	.	.	PUNCT
easat-9471	257	1	this	this	DET
easat-9471	257	2	open	open	ADJ
easat-9471	257	3	-	-	PUNCT
easat-9471	257	4	access	access	NOUN
easat-9471	257	5	article	article	NOUN
easat-9471	257	6	is	be	AUX
easat-9471	257	7	distributed	distribute	VERB
easat-9471	257	8	under	under	ADP
easat-9471	257	9	the	the	DET
easat-9471	257	10	terms	term	NOUN
easat-9471	257	11	and	and	CCONJ
easat-9471	257	12	conditions	condition	NOUN
easat-9471	257	13	of	of	ADP
easat-9471	257	14	the	the	DET
easat-9471	257	15	creative	creative	ADJ
easat-9471	257	16	commons	common	NOUN
easat-9471	257	17	attribution	attribution	NOUN
easat-9471	257	18	(	(	PUNCT
easat-9471	257	19	cc	cc	NOUN
easat-9471	257	20	by	by	ADP
easat-9471	257	21	)	)	PUNCT
easat-9471	257	22	license	license	NOUN
easat-9471	257	23	(	(	PUNCT
easat-9471	257	24	https://creativecommons.org/licenses/by/4.0/	https://creativecommons.org/licenses/by/4.0/	PROPN
easat-9471	257	25	)	)	PUNCT
easat-9471	257	26	.	.	PUNCT
easat-9471	258	1	https://creativecommons.org/licenses/by/4.0/	https://creativecommons.org/licenses/by/4.0/	PROPN
easat-9471	258	2	880	880	NUM
easat-9471	258	3	edelweiss	edelweiss	PROPN
easat-9471	258	4	applied	apply	VERB
easat-9471	258	5	science	science	NOUN
easat-9471	258	6	and	and	CCONJ
easat-9471	258	7	technology	technology	NOUN
easat-9471	258	8	issn	issn	PROPN
easat-9471	258	9	:	:	PUNCT
easat-9471	258	10	2576	2576	NUM
easat-9471	258	11	-	-	SYM
easat-9471	258	12	8484	8484	NUM
easat-9471	258	13	vol	vol	NOUN
easat-9471	258	14	.	.	PROPN
easat-9471	259	1	9	9	NUM
easat-9471	259	2	,	,	PUNCT
easat-9471	259	3	no	no	INTJ
easat-9471	259	4	.	.	NOUN
easat-9471	259	5	8	8	NUM
easat-9471	259	6	:	:	PUNCT
easat-9471	259	7	871	871	NUM
easat-9471	259	8	-	-	SYM
easat-9471	259	9	881	881	NUM
easat-9471	259	10	,	,	PUNCT
easat-9471	259	11	2025	2025	NUM
easat-9471	259	12	doi	doi	NOUN
easat-9471	259	13	:	:	PUNCT
easat-9471	259	14	10.55214/2576	10.55214/2576	NUM
easat-9471	259	15	-	-	PUNCT
easat-9471	259	16	8484.v9i8.9471	8484.v9i8.9471	NOUN
easat-9471	260	1	©	©	PROPN
easat-9471	260	2	2025	2025	NUM
easat-9471	260	3	by	by	ADP
easat-9471	260	4	the	the	DET
easat-9471	260	5	authors	author	NOUN
easat-9471	260	6	;	;	PUNCT
easat-9471	260	7	licensee	licensee	PROPN
easat-9471	260	8	learning	learn	VERB
easat-9471	260	9	gate	gate	NOUN
easat-9471	260	10	references	reference	NOUN
easat-9471	260	11	[	[	X
easat-9471	260	12	1	1	NUM
easat-9471	260	13	]	]	PUNCT
easat-9471	260	14	s.	s.	PROPN
easat-9471	260	15	rendle	rendle	PROPN
easat-9471	260	16	,	,	PUNCT
easat-9471	260	17	c.	c.	NOUN
easat-9471	260	18	freudenthaler	freudenthaler	NOUN
easat-9471	260	19	,	,	PUNCT
easat-9471	260	20	and	and	CCONJ
easat-9471	260	21	l.	l.	PROPN
easat-9471	260	22	schmidt	schmidt	PROPN
easat-9471	260	23	-	-	PUNCT
easat-9471	260	24	thieme	thieme	NOUN
easat-9471	260	25	,	,	PUNCT
easat-9471	260	26	"	"	PUNCT
easat-9471	260	27	factorizing	factorize	VERB
easat-9471	260	28	personalized	personalized	ADJ
easat-9471	260	29	markov	markov	NOUN
easat-9471	260	30	chains	chain	NOUN
easat-9471	260	31	for	for	ADP
easat-9471	260	32	next	next	ADJ
easat-9471	260	33	-	-	PUNCT
easat-9471	260	34	basket	basket	NOUN
easat-9471	260	35	recommendation	recommendation	NOUN
easat-9471	260	36	,	,	PUNCT
easat-9471	260	37	"	"	PUNCT
easat-9471	260	38	presented	present	VERB
easat-9471	260	39	at	at	ADP
easat-9471	260	40	the	the	DET
easat-9471	260	41	proceedings	proceeding	NOUN
easat-9471	260	42	of	of	ADP
easat-9471	260	43	the	the	DET
easat-9471	260	44	19th	19th	ADJ
easat-9471	260	45	international	international	ADJ
easat-9471	260	46	conference	conference	NOUN
easat-9471	260	47	on	on	ADP
easat-9471	260	48	world	world	NOUN
easat-9471	260	49	wide	wide	ADJ
easat-9471	260	50	web	web	NOUN
easat-9471	260	51	(	(	PUNCT
easat-9471	260	52	pp	pp	ADJ
easat-9471	260	53	.	.	PUNCT
easat-9471	260	54	811820	811820	NUM
easat-9471	260	55	)	)	PUNCT
easat-9471	260	56	,	,	PUNCT
easat-9471	260	57	2010	2010	NUM
easat-9471	260	58	.	.	PUNCT
easat-9471	261	1	[	[	X
easat-9471	261	2	2	2	NUM
easat-9471	261	3	]	]	X
easat-9471	261	4	b.	b.	PROPN
easat-9471	261	5	hidasi	hidasi	PROPN
easat-9471	261	6	,	,	PUNCT
easat-9471	261	7	a.	a.	NOUN
easat-9471	261	8	karatzoglou	karatzoglou	PROPN
easat-9471	261	9	,	,	PUNCT
easat-9471	261	10	l.	l.	PROPN
easat-9471	261	11	baltrunas	baltrunas	PROPN
easat-9471	261	12	,	,	PUNCT
easat-9471	261	13	and	and	CCONJ
easat-9471	261	14	d.	d.	PROPN
easat-9471	261	15	tikk	tikk	PROPN
easat-9471	261	16	,	,	PUNCT
easat-9471	261	17	"	"	PUNCT
easat-9471	261	18	session	session	NOUN
easat-9471	261	19	-	-	PUNCT
easat-9471	261	20	based	base	VERB
easat-9471	261	21	recommendations	recommendation	NOUN
easat-9471	261	22	with	with	ADP
easat-9471	261	23	recurrent	recurrent	ADJ
easat-9471	261	24	neural	neural	ADJ
easat-9471	261	25	networks	network	NOUN
easat-9471	261	26	,	,	PUNCT
easat-9471	261	27	"	"	PUNCT
easat-9471	261	28	arxiv	arxiv	PROPN
easat-9471	261	29	preprint	preprint	VERB
easat-9471	261	30	arxiv:1511.06939	arxiv:1511.06939	NOUN
easat-9471	261	31	,	,	PUNCT
easat-9471	261	32	2015	2015	NUM
easat-9471	261	33	.	.	PUNCT
easat-9471	262	1	[	[	X
easat-9471	262	2	3	3	X
easat-9471	262	3	]	]	X
easat-9471	262	4	w.	w.	PROPN
easat-9471	262	5	c.	c.	PROPN
easat-9471	262	6	kang	kang	PROPN
easat-9471	262	7	and	and	CCONJ
easat-9471	262	8	j.	j.	PROPN
easat-9471	262	9	mcauley	mcauley	PROPN
easat-9471	262	10	,	,	PUNCT
easat-9471	262	11	"	"	PUNCT
easat-9471	262	12	self	self	NOUN
easat-9471	262	13	-	-	PUNCT
easat-9471	262	14	attentive	attentive	ADJ
easat-9471	262	15	sequential	sequential	ADJ
easat-9471	262	16	recommendation	recommendation	NOUN
easat-9471	262	17	,	,	PUNCT
easat-9471	262	18	"	"	PUNCT
easat-9471	262	19	presented	present	VERB
easat-9471	262	20	at	at	ADP
easat-9471	262	21	the	the	DET
easat-9471	262	22	ieee	ieee	NOUN
easat-9471	262	23	international	international	PROPN
easat-9471	262	24	conference	conference	NOUN
easat-9471	262	25	on	on	ADP
easat-9471	262	26	data	datum	NOUN
easat-9471	262	27	mining	mining	NOUN
easat-9471	262	28	(	(	PUNCT
easat-9471	262	29	icdm	icdm	NOUN
easat-9471	262	30	)	)	PUNCT
easat-9471	262	31	(	(	PUNCT
easat-9471	262	32	pp	pp	X
easat-9471	262	33	.	.	PUNCT
easat-9471	262	34	197	197	NUM
easat-9471	262	35	-	-	SYM
easat-9471	262	36	206	206	NUM
easat-9471	262	37	)	)	PUNCT
easat-9471	262	38	.	.	PUNCT
easat-9471	263	1	ieee	ieee	NOUN
easat-9471	263	2	,	,	PUNCT
easat-9471	263	3	2018	2018	NUM
easat-9471	263	4	.	.	PUNCT
easat-9471	264	1	[	[	X
easat-9471	264	2	4	4	X
easat-9471	264	3	]	]	PUNCT
easat-9471	264	4	x.	x.	NOUN
easat-9471	264	5	wang	wang	PROPN
easat-9471	264	6	,	,	PUNCT
easat-9471	264	7	x.	x.	PROPN
easat-9471	264	8	he	he	PRON
easat-9471	264	9	,	,	PUNCT
easat-9471	264	10	m.	m.	PROPN
easat-9471	264	11	wang	wang	PROPN
easat-9471	264	12	,	,	PUNCT
easat-9471	264	13	f.	f.	PROPN
easat-9471	264	14	feng	feng	PROPN
easat-9471	264	15	,	,	PUNCT
easat-9471	264	16	and	and	CCONJ
easat-9471	264	17	t.-s	t.-	NOUN
easat-9471	264	18	.	.	PUNCT
easat-9471	265	1	chua	chua	PROPN
easat-9471	265	2	,	,	PUNCT
easat-9471	265	3	"	"	PUNCT
easat-9471	265	4	neural	neural	ADJ
easat-9471	265	5	graph	graph	NOUN
easat-9471	265	6	collaborative	collaborative	ADJ
easat-9471	265	7	filtering	filtering	NOUN
easat-9471	265	8	,	,	PUNCT
easat-9471	265	9	"	"	PUNCT
easat-9471	265	10	in	in	ADP
easat-9471	265	11	proceedings	proceeding	NOUN
easat-9471	265	12	of	of	ADP
easat-9471	265	13	the	the	DET
easat-9471	265	14	42nd	42nd	PROPN
easat-9471	265	15	international	international	PROPN
easat-9471	265	16	acm	acm	PROPN
easat-9471	265	17	sigir	sigir	PROPN
easat-9471	265	18	conference	conference	PROPN
easat-9471	265	19	on	on	ADP
easat-9471	265	20	research	research	NOUN
easat-9471	265	21	and	and	CCONJ
easat-9471	265	22	development	development	NOUN
easat-9471	265	23	in	in	ADP
easat-9471	265	24	information	information	NOUN
easat-9471	265	25	retrieval	retrieval	NOUN
easat-9471	265	26	,	,	PUNCT
easat-9471	265	27	2019	2019	NUM
easat-9471	265	28	.	.	PUNCT
easat-9471	266	1	[	[	X
easat-9471	266	2	5	5	NUM
easat-9471	266	3	]	]	PUNCT
easat-9471	266	4	x.	x.	NOUN
easat-9471	266	5	he	he	PRON
easat-9471	266	6	,	,	PUNCT
easat-9471	266	7	k.	k.	PROPN
easat-9471	266	8	deng	deng	PROPN
easat-9471	266	9	,	,	PUNCT
easat-9471	266	10	x.	x.	PROPN
easat-9471	266	11	wang	wang	PROPN
easat-9471	266	12	,	,	PUNCT
easat-9471	266	13	y.	y.	PROPN
easat-9471	266	14	li	li	PROPN
easat-9471	266	15	,	,	PUNCT
easat-9471	266	16	y.	y.	PROPN
easat-9471	266	17	zhang	zhang	PROPN
easat-9471	266	18	,	,	PUNCT
easat-9471	266	19	and	and	CCONJ
easat-9471	266	20	m.	m.	PROPN
easat-9471	266	21	wang	wang	PROPN
easat-9471	266	22	,	,	PUNCT
easat-9471	266	23	"	"	PUNCT
easat-9471	266	24	lightgcn	lightgcn	NOUN
easat-9471	266	25	:	:	PUNCT
easat-9471	266	26	simplifying	simplify	VERB
easat-9471	266	27	and	and	CCONJ
easat-9471	266	28	powering	power	VERB
easat-9471	266	29	graph	graph	NOUN
easat-9471	266	30	convolution	convolution	NOUN
easat-9471	266	31	network	network	NOUN
easat-9471	266	32	for	for	ADP
easat-9471	266	33	recommendation	recommendation	NOUN
easat-9471	266	34	,	,	PUNCT
easat-9471	266	35	"	"	PUNCT
easat-9471	266	36	in	in	ADP
easat-9471	266	37	proceedings	proceeding	NOUN
easat-9471	266	38	of	of	ADP
easat-9471	266	39	the	the	DET
easat-9471	266	40	43rd	43rd	PROPN
easat-9471	266	41	international	international	PROPN
easat-9471	266	42	acm	acm	PROPN
easat-9471	266	43	sigir	sigir	PROPN
easat-9471	266	44	conference	conference	PROPN
easat-9471	266	45	on	on	ADP
easat-9471	266	46	research	research	NOUN
easat-9471	266	47	and	and	CCONJ
easat-9471	266	48	development	development	NOUN
easat-9471	266	49	in	in	ADP
easat-9471	266	50	information	information	NOUN
easat-9471	266	51	retrieval	retrieval	NOUN
easat-9471	266	52	,	,	PUNCT
easat-9471	266	53	2020	2020	NUM
easat-9471	266	54	.	.	PUNCT
easat-9471	267	1	[	[	X
easat-9471	267	2	6	6	NUM
easat-9471	267	3	]	]	PUNCT
easat-9471	267	4	s.	s.	PROPN
easat-9471	267	5	wu	wu	PROPN
easat-9471	267	6	,	,	PUNCT
easat-9471	267	7	y.	y.	PROPN
easat-9471	267	8	tang	tang	PROPN
easat-9471	267	9	,	,	PUNCT
easat-9471	267	10	y.	y.	PROPN
easat-9471	267	11	zhu	zhu	PROPN
easat-9471	267	12	,	,	PUNCT
easat-9471	267	13	l.	l.	PROPN
easat-9471	267	14	wang	wang	PROPN
easat-9471	267	15	,	,	PUNCT
easat-9471	267	16	x.	x.	PROPN
easat-9471	267	17	xie	xie	PROPN
easat-9471	267	18	,	,	PUNCT
easat-9471	267	19	and	and	CCONJ
easat-9471	267	20	t.	t.	PROPN
easat-9471	267	21	tan	tan	PROPN
easat-9471	267	22	,	,	PUNCT
easat-9471	267	23	"	"	PUNCT
easat-9471	267	24	session	session	NOUN
easat-9471	267	25	-	-	PUNCT
easat-9471	267	26	based	base	VERB
easat-9471	267	27	recommendation	recommendation	NOUN
easat-9471	267	28	with	with	ADP
easat-9471	267	29	graph	graph	NOUN
easat-9471	267	30	neural	neural	ADJ
easat-9471	267	31	networks	network	NOUN
easat-9471	267	32	,	,	PUNCT
easat-9471	267	33	"	"	PUNCT
easat-9471	267	34	proceedings	proceeding	NOUN
easat-9471	267	35	of	of	ADP
easat-9471	267	36	the	the	DET
easat-9471	267	37	aaai	aaai	PROPN
easat-9471	267	38	conference	conference	NOUN
easat-9471	267	39	on	on	ADP
easat-9471	267	40	artificial	artificial	ADJ
easat-9471	267	41	intelligence	intelligence	NOUN
easat-9471	267	42	,	,	PUNCT
easat-9471	267	43	vol	vol	NOUN
easat-9471	267	44	.	.	PROPN
easat-9471	268	1	33	33	NUM
easat-9471	268	2	,	,	PUNCT
easat-9471	268	3	no	no	INTJ
easat-9471	268	4	.	.	NOUN
easat-9471	268	5	1	1	NUM
easat-9471	268	6	,	,	PUNCT
easat-9471	268	7	pp	pp	ADJ
easat-9471	268	8	.	.	PUNCT
easat-9471	269	1	346–353	346–353	NUM
easat-9471	269	2	,	,	PUNCT
easat-9471	269	3	2019	2019	NUM
easat-9471	269	4	.	.	PUNCT
easat-9471	270	1	[	[	X
easat-9471	270	2	7	7	X
easat-9471	270	3	]	]	PUNCT
easat-9471	270	4	j.	j.	PROPN
easat-9471	270	5	chang	chang	PROPN
easat-9471	270	6	et	et	PROPN
easat-9471	270	7	al	al	PROPN
easat-9471	270	8	,	,	PUNCT
easat-9471	270	9	"	"	PUNCT
easat-9471	270	10	sequential	sequential	ADJ
easat-9471	270	11	recommendation	recommendation	NOUN
easat-9471	270	12	with	with	ADP
easat-9471	270	13	graph	graph	NOUN
easat-9471	270	14	neural	neural	ADJ
easat-9471	270	15	networks	network	NOUN
easat-9471	270	16	,	,	PUNCT
easat-9471	270	17	"	"	PUNCT
easat-9471	270	18	in	in	ADP
easat-9471	270	19	proceedings	proceeding	NOUN
easat-9471	270	20	of	of	ADP
easat-9471	270	21	the	the	DET
easat-9471	270	22	44th	44th	ADJ
easat-9471	270	23	international	international	PROPN
easat-9471	270	24	acm	acm	PROPN
easat-9471	270	25	sigir	sigir	PROPN
easat-9471	270	26	conference	conference	PROPN
easat-9471	270	27	on	on	ADP
easat-9471	270	28	research	research	NOUN
easat-9471	270	29	and	and	CCONJ
easat-9471	270	30	development	development	NOUN
easat-9471	270	31	in	in	ADP
easat-9471	270	32	information	information	NOUN
easat-9471	270	33	retrieval	retrieval	NOUN
easat-9471	270	34	,	,	PUNCT
easat-9471	270	35	2021	2021	NUM
easat-9471	270	36	.	.	PUNCT
easat-9471	271	1	[	[	X
easat-9471	271	2	8	8	NUM
easat-9471	271	3	]	]	X
easat-9471	271	4	y.	y.	PROPN
easat-9471	271	5	cao	cao	PROPN
easat-9471	271	6	et	et	PROPN
easat-9471	271	7	al	al	PROPN
easat-9471	271	8	,	,	PUNCT
easat-9471	271	9	"	"	PUNCT
easat-9471	271	10	graph	graph	NOUN
easat-9471	271	11	-	-	PUNCT
easat-9471	271	12	sequential	sequential	ADJ
easat-9471	271	13	alignment	alignment	NOUN
easat-9471	271	14	and	and	CCONJ
easat-9471	271	15	uniformity	uniformity	NOUN
easat-9471	271	16	:	:	PUNCT
easat-9471	271	17	toward	toward	ADP
easat-9471	271	18	enhanced	enhanced	ADJ
easat-9471	271	19	recommendation	recommendation	NOUN
easat-9471	271	20	systems	system	NOUN
easat-9471	271	21	,	,	PUNCT
easat-9471	271	22	"	"	PUNCT
easat-9471	271	23	in	in	ADP
easat-9471	271	24	companion	companion	ADJ
easat-9471	271	25	proceedings	proceeding	NOUN
easat-9471	271	26	of	of	ADP
easat-9471	271	27	the	the	DET
easat-9471	271	28	acm	acm	NOUN
easat-9471	271	29	on	on	ADP
easat-9471	271	30	web	web	NOUN
easat-9471	271	31	conference	conference	NOUN
easat-9471	271	32	,	,	PUNCT
easat-9471	271	33	2025	2025	NUM
easat-9471	271	34	.	.	PUNCT
easat-9471	272	1	[	[	X
easat-9471	272	2	9	9	NUM
easat-9471	272	3	]	]	PUNCT
easat-9471	272	4	x.	x.	NOUN
easat-9471	272	5	zhao	zhao	PROPN
easat-9471	272	6	,	,	PUNCT
easat-9471	272	7	l.	l.	PROPN
easat-9471	272	8	xia	xia	PROPN
easat-9471	272	9	,	,	PUNCT
easat-9471	272	10	l.	l.	PROPN
easat-9471	272	11	zou	zou	PROPN
easat-9471	272	12	,	,	PUNCT
easat-9471	272	13	h.	h.	PROPN
easat-9471	272	14	liu	liu	PROPN
easat-9471	272	15	,	,	PUNCT
easat-9471	272	16	d.	d.	PROPN
easat-9471	272	17	yin	yin	PROPN
easat-9471	272	18	,	,	PUNCT
easat-9471	272	19	and	and	CCONJ
easat-9471	272	20	j.	j.	PROPN
easat-9471	272	21	tang	tang	PROPN
easat-9471	272	22	,	,	PUNCT
easat-9471	272	23	"	"	PUNCT
easat-9471	272	24	whole	whole	ADJ
easat-9471	272	25	-	-	PUNCT
easat-9471	272	26	chain	chain	NOUN
easat-9471	272	27	recommendations	recommendation	NOUN
easat-9471	272	28	,	,	PUNCT
easat-9471	272	29	"	"	PUNCT
easat-9471	272	30	in	in	ADP
easat-9471	272	31	proceedings	proceeding	NOUN
easat-9471	272	32	of	of	ADP
easat-9471	272	33	the	the	DET
easat-9471	272	34	29th	29th	ADJ
easat-9471	272	35	acm	acm	PROPN
easat-9471	272	36	international	international	ADJ
easat-9471	272	37	conference	conference	NOUN
easat-9471	272	38	on	on	ADP
easat-9471	272	39	information	information	NOUN
easat-9471	272	40	&	&	CCONJ
easat-9471	272	41	knowledge	knowledge	PROPN
easat-9471	272	42	management	management	PROPN
easat-9471	272	43	,	,	PUNCT
easat-9471	272	44	2020	2020	NUM
easat-9471	272	45	.	.	PUNCT
easat-9471	273	1	[	[	X
easat-9471	273	2	10	10	NUM
easat-9471	273	3	]	]	X
easat-9471	273	4	l.	l.	PROPN
easat-9471	273	5	zou	zou	PROPN
easat-9471	273	6	,	,	PUNCT
easat-9471	273	7	l.	l.	PROPN
easat-9471	273	8	xia	xia	PROPN
easat-9471	273	9	,	,	PUNCT
easat-9471	273	10	z.	z.	PROPN
easat-9471	273	11	ding	ding	PROPN
easat-9471	273	12	,	,	PUNCT
easat-9471	273	13	j.	j.	PROPN
easat-9471	273	14	song	song	PROPN
easat-9471	273	15	,	,	PUNCT
easat-9471	273	16	w.	w.	PROPN
easat-9471	273	17	liu	liu	PROPN
easat-9471	273	18	,	,	PUNCT
easat-9471	273	19	and	and	CCONJ
easat-9471	273	20	d.	d.	PROPN
easat-9471	273	21	yin	yin	PROPN
easat-9471	273	22	,	,	PUNCT
easat-9471	273	23	"	"	PUNCT
easat-9471	273	24	reinforcement	reinforcement	NOUN
easat-9471	273	25	learning	learning	NOUN
easat-9471	273	26	to	to	PART
easat-9471	273	27	optimize	optimize	VERB
easat-9471	273	28	long	long	ADJ
easat-9471	273	29	-	-	PUNCT
easat-9471	273	30	term	term	NOUN
easat-9471	273	31	user	user	NOUN
easat-9471	273	32	engagement	engagement	NOUN
easat-9471	273	33	in	in	ADP
easat-9471	273	34	recommender	recommender	NOUN
easat-9471	273	35	systems	system	NOUN
easat-9471	273	36	,	,	PUNCT
easat-9471	273	37	"	"	PUNCT
easat-9471	273	38	in	in	ADP
easat-9471	273	39	proceedings	proceeding	NOUN
easat-9471	273	40	of	of	ADP
easat-9471	273	41	the	the	DET
easat-9471	273	42	25th	25th	ADJ
easat-9471	273	43	acm	acm	PROPN
easat-9471	273	44	sigkdd	sigkdd	NOUN
easat-9471	273	45	international	international	ADJ
easat-9471	273	46	conference	conference	NOUN
easat-9471	273	47	on	on	ADP
easat-9471	273	48	knowledge	knowledge	PROPN
easat-9471	273	49	discovery	discovery	PROPN
easat-9471	273	50	&	&	CCONJ
easat-9471	273	51	data	data	PROPN
easat-9471	273	52	mining	mining	NOUN
easat-9471	273	53	,	,	PUNCT
easat-9471	273	54	2019	2019	NUM
easat-9471	273	55	.	.	PUNCT
easat-9471	274	1	[	[	X
easat-9471	274	2	11	11	NUM
easat-9471	274	3	]	]	PUNCT
easat-9471	274	4	m.	m.	NOUN
easat-9471	274	5	chen	chen	PROPN
easat-9471	274	6	,	,	PUNCT
easat-9471	274	7	a.	a.	NOUN
easat-9471	274	8	beutel	beutel	PROPN
easat-9471	274	9	,	,	PUNCT
easat-9471	274	10	p.	p.	PROPN
easat-9471	274	11	covington	covington	PROPN
easat-9471	274	12	,	,	PUNCT
easat-9471	274	13	s.	s.	PROPN
easat-9471	274	14	jain	jain	PROPN
easat-9471	274	15	,	,	PUNCT
easat-9471	274	16	f.	f.	PROPN
easat-9471	274	17	belletti	belletti	PROPN
easat-9471	274	18	,	,	PUNCT
easat-9471	274	19	and	and	CCONJ
easat-9471	274	20	e.	e.	PROPN
easat-9471	274	21	h.	h.	PROPN
easat-9471	274	22	chi	chi	PROPN
easat-9471	274	23	,	,	PUNCT
easat-9471	274	24	"	"	PUNCT
easat-9471	274	25	top	top	ADJ
easat-9471	274	26	-	-	PUNCT
easat-9471	274	27	k	k	ADJ
easat-9471	274	28	off	off	ADP
easat-9471	274	29	-	-	PUNCT
easat-9471	274	30	policy	policy	NOUN
easat-9471	274	31	correction	correction	NOUN
easat-9471	274	32	for	for	ADP
easat-9471	274	33	a	a	DET
easat-9471	274	34	reinforce	reinforce	NOUN
easat-9471	274	35	recommender	recommender	NOUN
easat-9471	274	36	system	system	NOUN
easat-9471	274	37	,	,	PUNCT
easat-9471	274	38	"	"	PUNCT
easat-9471	274	39	in	in	ADP
easat-9471	274	40	proceedings	proceeding	NOUN
easat-9471	274	41	of	of	ADP
easat-9471	274	42	the	the	DET
easat-9471	274	43	twelfth	twelfth	ADJ
easat-9471	274	44	acm	acm	PROPN
easat-9471	274	45	international	international	PROPN
easat-9471	274	46	conference	conference	NOUN
easat-9471	274	47	on	on	ADP
easat-9471	274	48	web	web	NOUN
easat-9471	274	49	search	search	NOUN
easat-9471	274	50	and	and	CCONJ
easat-9471	274	51	data	datum	NOUN
easat-9471	274	52	mining	mining	NOUN
easat-9471	274	53	,	,	PUNCT
easat-9471	274	54	2019	2019	NUM
easat-9471	274	55	.	.	PUNCT
easat-9471	275	1	[	[	X
easat-9471	275	2	12	12	NUM
easat-9471	275	3	]	]	PUNCT
easat-9471	275	4	m.	m.	NOUN
easat-9471	275	5	chen	chen	PROPN
easat-9471	275	6	,	,	PUNCT
easat-9471	275	7	c.	c.	PROPN
easat-9471	275	8	xu	xu	PROPN
easat-9471	275	9	,	,	PUNCT
easat-9471	275	10	v.	v.	CCONJ
easat-9471	275	11	gatto	gatto	PROPN
easat-9471	275	12	,	,	PUNCT
easat-9471	275	13	d.	d.	PROPN
easat-9471	275	14	jain	jain	PROPN
easat-9471	275	15	,	,	PUNCT
easat-9471	275	16	a.	a.	PROPN
easat-9471	275	17	kumar	kumar	PROPN
easat-9471	275	18	,	,	PUNCT
easat-9471	275	19	and	and	CCONJ
easat-9471	275	20	e.	e.	PROPN
easat-9471	275	21	chi	chi	PROPN
easat-9471	275	22	,	,	PUNCT
easat-9471	275	23	"	"	PUNCT
easat-9471	275	24	off	off	ADP
easat-9471	275	25	-	-	PUNCT
easat-9471	275	26	policy	policy	NOUN
easat-9471	275	27	actor	actor	NOUN
easat-9471	275	28	-	-	PUNCT
easat-9471	275	29	critic	critic	NOUN
easat-9471	275	30	for	for	ADP
easat-9471	275	31	recommender	recommender	NOUN
easat-9471	275	32	systems	system	NOUN
easat-9471	275	33	,	,	PUNCT
easat-9471	275	34	"	"	PUNCT
easat-9471	275	35	in	in	ADP
easat-9471	275	36	proceedings	proceeding	NOUN
easat-9471	275	37	of	of	ADP
easat-9471	275	38	the	the	DET
easat-9471	275	39	16th	16th	ADJ
easat-9471	275	40	acm	acm	NOUN
easat-9471	275	41	conference	conference	NOUN
easat-9471	275	42	on	on	ADP
easat-9471	275	43	recommender	recommender	NOUN
easat-9471	275	44	systems	system	NOUN
easat-9471	275	45	,	,	PUNCT
easat-9471	275	46	2022	2022	NUM
easat-9471	275	47	.	.	PUNCT
easat-9471	276	1	[	[	X
easat-9471	276	2	13	13	NUM
easat-9471	276	3	]	]	X
easat-9471	276	4	d.	d.	PROPN
easat-9471	276	5	guo	guo	PROPN
easat-9471	276	6	et	et	PROPN
easat-9471	276	7	al	al	PROPN
easat-9471	276	8	.	.	PROPN
easat-9471	276	9	,	,	PUNCT
easat-9471	276	10	"	"	PUNCT
easat-9471	276	11	deepseek	deepseek	ADJ
easat-9471	276	12	-	-	PUNCT
easat-9471	276	13	r1	r1	NOUN
easat-9471	276	14	:	:	PUNCT
easat-9471	276	15	incentivizing	incentivize	VERB
easat-9471	276	16	reasoning	reasoning	NOUN
easat-9471	276	17	capability	capability	NOUN
easat-9471	276	18	in	in	ADP
easat-9471	276	19	llms	llm	NOUN
easat-9471	276	20	via	via	ADP
easat-9471	276	21	reinforcement	reinforcement	NOUN
easat-9471	276	22	learning	learning	NOUN
easat-9471	276	23	,	,	PUNCT
easat-9471	276	24	"	"	PUNCT
easat-9471	276	25	arxiv	arxiv	PROPN
easat-9471	276	26	preprint	preprint	NOUN
easat-9471	276	27	arxiv:2501.12948	arxiv:2501.12948	NOUN
easat-9471	276	28	,	,	PUNCT
easat-9471	276	29	2025	2025	NUM
easat-9471	276	30	.	.	PUNCT
easat-9471	277	1	[	[	X
easat-9471	277	2	14	14	NUM
easat-9471	277	3	]	]	X
easat-9471	277	4	r.	r.	PROPN
easat-9471	277	5	qiu	qiu	PROPN
easat-9471	277	6	,	,	PUNCT
easat-9471	277	7	j.	j.	PROPN
easat-9471	277	8	li	li	PROPN
easat-9471	277	9	,	,	PUNCT
easat-9471	277	10	z.	z.	PROPN
easat-9471	277	11	huang	huang	PROPN
easat-9471	277	12	,	,	PUNCT
easat-9471	277	13	and	and	CCONJ
easat-9471	277	14	h.	h.	PROPN
easat-9471	277	15	yin	yin	PROPN
easat-9471	277	16	,	,	PUNCT
easat-9471	277	17	"	"	PUNCT
easat-9471	277	18	rethinking	rethink	VERB
easat-9471	277	19	the	the	DET
easat-9471	277	20	item	item	NOUN
easat-9471	277	21	order	order	NOUN
easat-9471	277	22	in	in	ADP
easat-9471	277	23	session	session	NOUN
easat-9471	277	24	-	-	PUNCT
easat-9471	277	25	based	base	VERB
easat-9471	277	26	recommendation	recommendation	NOUN
easat-9471	277	27	with	with	ADP
easat-9471	277	28	graph	graph	NOUN
easat-9471	277	29	neural	neural	ADJ
easat-9471	277	30	networks	network	NOUN
easat-9471	277	31	,	,	PUNCT
easat-9471	277	32	"	"	PUNCT
easat-9471	277	33	in	in	ADP
easat-9471	277	34	proceedings	proceeding	NOUN
easat-9471	277	35	of	of	ADP
easat-9471	277	36	the	the	DET
easat-9471	277	37	28th	28th	ADJ
easat-9471	277	38	acm	acm	PROPN
easat-9471	277	39	international	international	ADJ
easat-9471	277	40	conference	conference	NOUN
easat-9471	277	41	on	on	ADP
easat-9471	277	42	information	information	NOUN
easat-9471	277	43	and	and	CCONJ
easat-9471	277	44	knowledge	knowledge	NOUN
easat-9471	277	45	management	management	NOUN
easat-9471	277	46	,	,	PUNCT
easat-9471	277	47	2019	2019	NUM
easat-9471	277	48	.	.	PUNCT
easat-9471	278	1	[	[	X
easat-9471	278	2	15	15	NUM
easat-9471	278	3	]	]	X
easat-9471	278	4	f.	f.	PROPN
easat-9471	278	5	yu	yu	PROPN
easat-9471	278	6	,	,	PUNCT
easat-9471	278	7	y.	y.	PROPN
easat-9471	278	8	zhu	zhu	PROPN
easat-9471	278	9	,	,	PUNCT
easat-9471	278	10	q.	q.	PROPN
easat-9471	278	11	liu	liu	PROPN
easat-9471	278	12	,	,	PUNCT
easat-9471	278	13	s.	s.	PROPN
easat-9471	278	14	wu	wu	PROPN
easat-9471	278	15	,	,	PUNCT
easat-9471	278	16	l.	l.	PROPN
easat-9471	278	17	wang	wang	PROPN
easat-9471	278	18	,	,	PUNCT
easat-9471	278	19	and	and	CCONJ
easat-9471	278	20	t.	t.	PROPN
easat-9471	278	21	tan	tan	PROPN
easat-9471	278	22	,	,	PUNCT
easat-9471	278	23	"	"	PUNCT
easat-9471	278	24	tagnn	tagnn	NOUN
easat-9471	278	25	:	:	PUNCT
easat-9471	278	26	target	target	NOUN
easat-9471	278	27	attentive	attentive	ADJ
easat-9471	278	28	graph	graph	NOUN
easat-9471	278	29	neural	neural	ADJ
easat-9471	278	30	networks	network	NOUN
easat-9471	278	31	for	for	ADP
easat-9471	278	32	sessionbased	sessionbase	VERB
easat-9471	278	33	recommendation	recommendation	NOUN
easat-9471	278	34	,	,	PUNCT
easat-9471	278	35	"	"	PUNCT
easat-9471	278	36	in	in	ADP
easat-9471	278	37	proceedings	proceeding	NOUN
easat-9471	278	38	of	of	ADP
easat-9471	278	39	the	the	DET
easat-9471	278	40	43rd	43rd	PROPN
easat-9471	278	41	international	international	PROPN
easat-9471	278	42	acm	acm	PROPN
easat-9471	278	43	sigir	sigir	PROPN
easat-9471	278	44	conference	conference	PROPN
easat-9471	278	45	on	on	ADP
easat-9471	278	46	research	research	NOUN
easat-9471	278	47	and	and	CCONJ
easat-9471	278	48	development	development	NOUN
easat-9471	278	49	in	in	ADP
easat-9471	278	50	information	information	NOUN
easat-9471	278	51	retrieval	retrieval	NOUN
easat-9471	278	52	,	,	PUNCT
easat-9471	278	53	2020	2020	NUM
easat-9471	278	54	.	.	PUNCT
easat-9471	279	1	[	[	X
easat-9471	279	2	16	16	NUM
easat-9471	279	3	]	]	PUNCT
easat-9471	279	4	x.	x.	NOUN
easat-9471	279	5	xia	xia	PROPN
easat-9471	279	6	,	,	PUNCT
easat-9471	279	7	h.	h.	PROPN
easat-9471	279	8	yin	yin	PROPN
easat-9471	279	9	,	,	PUNCT
easat-9471	279	10	j.	j.	PROPN
easat-9471	279	11	yu	yu	PROPN
easat-9471	279	12	,	,	PUNCT
easat-9471	279	13	y.	y.	PROPN
easat-9471	279	14	shao	shao	PROPN
easat-9471	279	15	,	,	PUNCT
easat-9471	279	16	and	and	CCONJ
easat-9471	279	17	l.	l.	PROPN
easat-9471	279	18	cui	cui	PROPN
easat-9471	279	19	,	,	PUNCT
easat-9471	279	20	"	"	PUNCT
easat-9471	279	21	self	self	NOUN
easat-9471	279	22	-	-	PUNCT
easat-9471	279	23	supervised	supervise	VERB
easat-9471	279	24	graph	graph	NOUN
easat-9471	279	25	co	co	NOUN
easat-9471	279	26	-	-	NOUN
easat-9471	279	27	training	training	NOUN
easat-9471	279	28	for	for	ADP
easat-9471	279	29	session	session	NOUN
easat-9471	279	30	-	-	PUNCT
easat-9471	279	31	based	base	VERB
easat-9471	279	32	recommendation	recommendation	NOUN
easat-9471	279	33	,	,	PUNCT
easat-9471	279	34	"	"	PUNCT
easat-9471	279	35	in	in	ADP
easat-9471	279	36	proceedings	proceeding	NOUN
easat-9471	279	37	of	of	ADP
easat-9471	279	38	the	the	DET
easat-9471	279	39	30th	30th	ADJ
easat-9471	279	40	acm	acm	PROPN
easat-9471	279	41	international	international	ADJ
easat-9471	279	42	conference	conference	NOUN
easat-9471	279	43	on	on	ADP
easat-9471	279	44	information	information	NOUN
easat-9471	279	45	&	&	CCONJ
easat-9471	279	46	knowledge	knowledge	PROPN
easat-9471	279	47	management	management	PROPN
easat-9471	279	48	,	,	PUNCT
easat-9471	279	49	2021	2021	NUM
easat-9471	279	50	,	,	PUNCT
easat-9471	279	51	pp	pp	ADV
easat-9471	279	52	.	.	PUNCT
easat-9471	280	1	2180	2180	NUM
easat-9471	280	2	-	-	SYM
easat-9471	280	3	2190	2190	NUM
easat-9471	280	4	.	.	PUNCT
easat-9471	281	1	[	[	X
easat-9471	281	2	17	17	NUM
easat-9471	281	3	]	]	PUNCT
easat-9471	281	4	z.	z.	PROPN
easat-9471	281	5	wang	wang	PROPN
easat-9471	281	6	,	,	PUNCT
easat-9471	281	7	w.	w.	PROPN
easat-9471	281	8	wei	wei	PROPN
easat-9471	281	9	,	,	PUNCT
easat-9471	281	10	g.	g.	PROPN
easat-9471	281	11	cong	cong	PROPN
easat-9471	281	12	,	,	PUNCT
easat-9471	281	13	x.-l	x.-l	PROPN
easat-9471	281	14	.	.	PUNCT
easat-9471	282	1	li	li	PROPN
easat-9471	282	2	,	,	PUNCT
easat-9471	282	3	x.-l	x.-l	PROPN
easat-9471	282	4	.	.	PUNCT
easat-9471	283	1	mao	mao	PROPN
easat-9471	283	2	,	,	PUNCT
easat-9471	283	3	and	and	CCONJ
easat-9471	283	4	m.	m.	PROPN
easat-9471	283	5	qiu	qiu	PROPN
easat-9471	283	6	,	,	PUNCT
easat-9471	283	7	"	"	PUNCT
easat-9471	283	8	global	global	ADJ
easat-9471	283	9	context	context	NOUN
easat-9471	283	10	enhanced	enhance	VERB
easat-9471	283	11	graph	graph	NOUN
easat-9471	283	12	neural	neural	ADJ
easat-9471	283	13	networks	network	NOUN
easat-9471	283	14	for	for	ADP
easat-9471	283	15	session	session	NOUN
easat-9471	283	16	-	-	PUNCT
easat-9471	283	17	based	base	VERB
easat-9471	283	18	recommendation	recommendation	NOUN
easat-9471	283	19	,	,	PUNCT
easat-9471	283	20	"	"	PUNCT
easat-9471	283	21	in	in	ADP
easat-9471	283	22	proceedings	proceeding	NOUN
easat-9471	283	23	of	of	ADP
easat-9471	283	24	the	the	DET
easat-9471	283	25	43rd	43rd	PROPN
easat-9471	283	26	international	international	PROPN
easat-9471	283	27	acm	acm	PROPN
easat-9471	283	28	sigir	sigir	PROPN
easat-9471	283	29	conference	conference	PROPN
easat-9471	283	30	on	on	ADP
easat-9471	283	31	research	research	NOUN
easat-9471	283	32	and	and	CCONJ
easat-9471	283	33	development	development	NOUN
easat-9471	283	34	in	in	ADP
easat-9471	283	35	information	information	NOUN
easat-9471	283	36	retrieval	retrieval	NOUN
easat-9471	283	37	,	,	PUNCT
easat-9471	283	38	2020	2020	NUM
easat-9471	283	39	,	,	PUNCT
easat-9471	283	40	pp	pp	ADP
easat-9471	283	41	.	.	PUNCT
easat-9471	284	1	169	169	NUM
easat-9471	284	2	-	-	SYM
easat-9471	284	3	178	178	NUM
easat-9471	284	4	.	.	PUNCT
easat-9471	285	1	[	[	X
easat-9471	285	2	18	18	NUM
easat-9471	285	3	]	]	X
easat-9471	285	4	r.	r.	PROPN
easat-9471	285	5	qiu	qiu	PROPN
easat-9471	285	6	,	,	PUNCT
easat-9471	285	7	h.	h.	PROPN
easat-9471	285	8	yin	yin	PROPN
easat-9471	285	9	,	,	PUNCT
easat-9471	285	10	z.	z.	PROPN
easat-9471	285	11	huang	huang	PROPN
easat-9471	285	12	,	,	PUNCT
easat-9471	285	13	and	and	CCONJ
easat-9471	285	14	t.	t.	PROPN
easat-9471	285	15	chen	chen	PROPN
easat-9471	285	16	,	,	PUNCT
easat-9471	285	17	"	"	PUNCT
easat-9471	285	18	gag	gag	NOUN
easat-9471	285	19	:	:	PUNCT
easat-9471	285	20	global	global	ADJ
easat-9471	285	21	attributed	attribute	VERB
easat-9471	285	22	graph	graph	NOUN
easat-9471	285	23	neural	neural	ADJ
easat-9471	285	24	network	network	NOUN
easat-9471	285	25	for	for	ADP
easat-9471	285	26	streaming	streaming	NOUN
easat-9471	285	27	session	session	NOUN
easat-9471	285	28	-	-	PUNCT
easat-9471	285	29	based	base	VERB
easat-9471	285	30	recommendation	recommendation	NOUN
easat-9471	285	31	,	,	PUNCT
easat-9471	285	32	"	"	PUNCT
easat-9471	285	33	in	in	ADP
easat-9471	285	34	proceedings	proceeding	NOUN
easat-9471	285	35	of	of	ADP
easat-9471	285	36	the	the	DET
easat-9471	285	37	43rd	43rd	PROPN
easat-9471	285	38	international	international	PROPN
easat-9471	285	39	acm	acm	PROPN
easat-9471	285	40	sigir	sigir	PROPN
easat-9471	285	41	conference	conference	PROPN
easat-9471	285	42	on	on	ADP
easat-9471	285	43	research	research	NOUN
easat-9471	285	44	and	and	CCONJ
easat-9471	285	45	development	development	NOUN
easat-9471	285	46	in	in	ADP
easat-9471	285	47	information	information	NOUN
easat-9471	285	48	retrieval	retrieval	NOUN
easat-9471	285	49	,	,	PUNCT
easat-9471	285	50	2020	2020	NUM
easat-9471	285	51	,	,	PUNCT
easat-9471	285	52	pp	pp	ADV
easat-9471	285	53	.	.	PUNCT
easat-9471	286	1	669	669	NUM
easat-9471	286	2	-	-	SYM
easat-9471	286	3	678	678	NUM
easat-9471	286	4	.	.	PUNCT
easat-9471	287	1	[	[	X
easat-9471	287	2	19	19	NUM
easat-9471	287	3	]	]	X
easat-9471	287	4	c.	c.	PROPN
easat-9471	287	5	j.	j.	PROPN
easat-9471	287	6	watkins	watkins	PROPN
easat-9471	287	7	and	and	CCONJ
easat-9471	287	8	p.	p.	PROPN
easat-9471	287	9	dayan	dayan	PROPN
easat-9471	287	10	,	,	PUNCT
easat-9471	287	11	"	"	PUNCT
easat-9471	287	12	q	q	X
easat-9471	287	13	-	-	PUNCT
easat-9471	287	14	learning	learning	NOUN
easat-9471	287	15	,	,	PUNCT
easat-9471	287	16	"	"	PUNCT
easat-9471	287	17	machine	machine	NOUN
easat-9471	287	18	learning	learning	NOUN
easat-9471	287	19	,	,	PUNCT
easat-9471	287	20	vol	vol	NOUN
easat-9471	287	21	.	.	PROPN
easat-9471	287	22	8	8	NUM
easat-9471	287	23	,	,	PUNCT
easat-9471	287	24	no	no	INTJ
easat-9471	287	25	.	.	NOUN
easat-9471	287	26	3	3	NUM
easat-9471	287	27	,	,	PUNCT
easat-9471	287	28	pp	pp	ADJ
easat-9471	287	29	.	.	PUNCT
easat-9471	287	30	279	279	NUM
easat-9471	287	31	-	-	SYM
easat-9471	287	32	292	292	NUM
easat-9471	287	33	,	,	PUNCT
easat-9471	287	34	1992	1992	NUM
easat-9471	287	35	.	.	PUNCT
easat-9471	288	1	[	[	X
easat-9471	288	2	20	20	NUM
easat-9471	288	3	]	]	X
easat-9471	288	4	v.	v.	PROPN
easat-9471	288	5	mnih	mnih	PROPN
easat-9471	288	6	et	et	PROPN
easat-9471	288	7	al	al	PROPN
easat-9471	288	8	.	.	PROPN
easat-9471	288	9	,	,	PUNCT
easat-9471	288	10	"	"	PUNCT
easat-9471	288	11	playing	play	VERB
easat-9471	288	12	atari	atari	NOUN
easat-9471	288	13	with	with	ADP
easat-9471	288	14	deep	deep	ADJ
easat-9471	288	15	reinforcement	reinforcement	NOUN
easat-9471	288	16	learning	learning	NOUN
easat-9471	288	17	,	,	PUNCT
easat-9471	288	18	"	"	PUNCT
easat-9471	288	19	arxiv	arxiv	PROPN
easat-9471	288	20	preprint	preprint	NOUN
easat-9471	288	21	arxiv:1312.5602	arxiv:1312.5602	NOUN
easat-9471	288	22	,	,	PUNCT
easat-9471	288	23	2013	2013	NUM
easat-9471	288	24	.	.	PUNCT
easat-9471	289	1	[	[	X
easat-9471	289	2	21	21	NUM
easat-9471	289	3	]	]	X
easat-9471	289	4	r.	r.	PROPN
easat-9471	289	5	s.	s.	PROPN
easat-9471	289	6	sutton	sutton	PROPN
easat-9471	289	7	,	,	PUNCT
easat-9471	289	8	d.	d.	PROPN
easat-9471	289	9	a.	a.	PROPN
easat-9471	289	10	mcallester	mcallester	PROPN
easat-9471	289	11	,	,	PUNCT
easat-9471	289	12	s.	s.	PROPN
easat-9471	289	13	p.	p.	PROPN
easat-9471	289	14	singh	singh	PROPN
easat-9471	289	15	,	,	PUNCT
easat-9471	289	16	and	and	CCONJ
easat-9471	289	17	y.	y.	PROPN
easat-9471	289	18	mansour	mansour	PROPN
easat-9471	289	19	,	,	PUNCT
easat-9471	289	20	"	"	PUNCT
easat-9471	289	21	policy	policy	NOUN
easat-9471	289	22	gradient	gradient	NOUN
easat-9471	289	23	methods	method	NOUN
easat-9471	289	24	for	for	ADP
easat-9471	289	25	reinforcement	reinforcement	NOUN
easat-9471	289	26	learning	learning	NOUN
easat-9471	289	27	with	with	ADP
easat-9471	289	28	function	function	NOUN
easat-9471	289	29	approximation	approximation	NOUN
easat-9471	289	30	,	,	PUNCT
easat-9471	289	31	"	"	PUNCT
easat-9471	289	32	advances	advance	NOUN
easat-9471	289	33	in	in	ADP
easat-9471	289	34	neural	neural	ADJ
easat-9471	289	35	information	information	NOUN
easat-9471	289	36	processing	processing	NOUN
easat-9471	289	37	systems	system	NOUN
easat-9471	289	38	,	,	PUNCT
easat-9471	289	39	vol	vol	NOUN
easat-9471	289	40	.	.	PROPN
easat-9471	289	41	12	12	NUM
easat-9471	289	42	,	,	PUNCT
easat-9471	289	43	pp	pp	PROPN
easat-9471	289	44	.	.	PUNCT
easat-9471	290	1	1057–106	1057–106	NUM
easat-9471	290	2	,	,	PUNCT
easat-9471	290	3	1999	1999	NUM
easat-9471	290	4	.	.	PUNCT
easat-9471	291	1	[	[	X
easat-9471	291	2	22	22	NUM
easat-9471	291	3	]	]	PUNCT
easat-9471	291	4	j.	j.	PROPN
easat-9471	291	5	schulman	schulman	PROPN
easat-9471	291	6	,	,	PUNCT
easat-9471	291	7	s.	s.	PROPN
easat-9471	291	8	levine	levine	PROPN
easat-9471	291	9	,	,	PUNCT
easat-9471	291	10	p.	p.	PROPN
easat-9471	291	11	abbeel	abbeel	PROPN
easat-9471	291	12	,	,	PUNCT
easat-9471	291	13	m.	m.	PROPN
easat-9471	291	14	jordan	jordan	PROPN
easat-9471	291	15	,	,	PUNCT
easat-9471	291	16	and	and	CCONJ
easat-9471	291	17	p.	p.	PROPN
easat-9471	291	18	moritz	moritz	PROPN
easat-9471	291	19	,	,	PUNCT
easat-9471	291	20	"	"	PUNCT
easat-9471	291	21	trust	trust	NOUN
easat-9471	291	22	region	region	NOUN
easat-9471	291	23	policy	policy	NOUN
easat-9471	291	24	optimization	optimization	NOUN
easat-9471	291	25	,	,	PUNCT
easat-9471	291	26	"	"	PUNCT
easat-9471	291	27	in	in	ADP
easat-9471	291	28	international	international	ADJ
easat-9471	291	29	conference	conference	NOUN
easat-9471	291	30	on	on	ADP
easat-9471	291	31	machine	machine	NOUN
easat-9471	291	32	learning	learning	NOUN
easat-9471	291	33	,	,	PUNCT
easat-9471	291	34	2015	2015	NUM
easat-9471	291	35	:	:	PUNCT
easat-9471	291	36	pmlr	pmlr	NOUN
easat-9471	291	37	,	,	PUNCT
easat-9471	291	38	pp	pp	ADJ
easat-9471	291	39	.	.	PUNCT
easat-9471	292	1	1889	1889	NUM
easat-9471	292	2	-	-	SYM
easat-9471	292	3	1897	1897	NUM
easat-9471	292	4	.	.	PUNCT
easat-9471	293	1	[	[	X
easat-9471	293	2	23	23	NUM
easat-9471	293	3	]	]	PUNCT
easat-9471	293	4	j.	j.	PROPN
easat-9471	293	5	schulman	schulman	PROPN
easat-9471	293	6	,	,	PUNCT
easat-9471	293	7	f.	f.	PROPN
easat-9471	293	8	wolski	wolski	PROPN
easat-9471	293	9	,	,	PUNCT
easat-9471	293	10	p.	p.	NOUN
easat-9471	293	11	dhariwal	dhariwal	NOUN
easat-9471	293	12	,	,	PUNCT
easat-9471	293	13	a.	a.	PROPN
easat-9471	293	14	radford	radford	PROPN
easat-9471	293	15	,	,	PUNCT
easat-9471	293	16	and	and	CCONJ
easat-9471	293	17	o.	o.	PROPN
easat-9471	293	18	klimov	klimov	PROPN
easat-9471	293	19	,	,	PUNCT
easat-9471	293	20	"	"	PUNCT
easat-9471	293	21	proximal	proximal	ADJ
easat-9471	293	22	policy	policy	NOUN
easat-9471	293	23	optimization	optimization	NOUN
easat-9471	293	24	algorithms	algorithm	NOUN
easat-9471	293	25	,	,	PUNCT
easat-9471	293	26	"	"	PUNCT
easat-9471	293	27	arxiv	arxiv	PROPN
easat-9471	293	28	preprint	preprint	NOUN
easat-9471	293	29	arxiv:1707.06347	arxiv:1707.06347	PROPN
easat-9471	293	30	,	,	PUNCT
easat-9471	293	31	2017	2017	NUM
easat-9471	293	32	.	.	PUNCT
easat-9471	294	1	[	[	X
easat-9471	294	2	24	24	NUM
easat-9471	294	3	]	]	PUNCT
easat-9471	294	4	l.	l.	PROPN
easat-9471	294	5	ouyang	ouyang	PROPN
easat-9471	294	6	et	et	PROPN
easat-9471	294	7	al	al	PROPN
easat-9471	294	8	.	.	PROPN
easat-9471	294	9	,	,	PUNCT
easat-9471	294	10	"	"	PUNCT
easat-9471	294	11	training	training	NOUN
easat-9471	294	12	language	language	NOUN
easat-9471	294	13	models	model	NOUN
easat-9471	294	14	to	to	PART
easat-9471	294	15	follow	follow	VERB
easat-9471	294	16	instructions	instruction	NOUN
easat-9471	294	17	with	with	ADP
easat-9471	294	18	human	human	ADJ
easat-9471	294	19	feedback	feedback	NOUN
easat-9471	294	20	,	,	PUNCT
easat-9471	294	21	"	"	PUNCT
easat-9471	294	22	advances	advance	NOUN
easat-9471	294	23	in	in	ADP
easat-9471	294	24	neural	neural	ADJ
easat-9471	294	25	information	information	NOUN
easat-9471	294	26	processing	processing	NOUN
easat-9471	294	27	systems	system	NOUN
easat-9471	294	28	,	,	PUNCT
easat-9471	294	29	vol	vol	NOUN
easat-9471	294	30	.	.	PROPN
easat-9471	294	31	35	35	NUM
easat-9471	294	32	,	,	PUNCT
easat-9471	294	33	pp	pp	ADJ
easat-9471	294	34	.	.	PUNCT
easat-9471	295	1	27730–27744	27730–27744	NUM
easat-9471	295	2	,	,	PUNCT
easat-9471	295	3	2002	2002	NUM
easat-9471	295	4	.	.	PUNCT
easat-9471	296	1	[	[	X
easat-9471	296	2	25	25	NUM
easat-9471	296	3	]	]	X
easat-9471	296	4	d.	d.	PROPN
easat-9471	296	5	yu	yu	PROPN
easat-9471	296	6	et	et	PROPN
easat-9471	296	7	al	al	PROPN
easat-9471	296	8	.	.	PROPN
easat-9471	296	9	,	,	PUNCT
easat-9471	296	10	"	"	PUNCT
easat-9471	296	11	crossing	cross	VERB
easat-9471	296	12	the	the	DET
easat-9471	296	13	reward	reward	NOUN
easat-9471	296	14	bridge	bridge	NOUN
easat-9471	296	15	:	:	PUNCT
easat-9471	296	16	expanding	expand	VERB
easat-9471	296	17	rl	rl	PRON
easat-9471	296	18	with	with	ADP
easat-9471	296	19	verifiable	verifiable	ADJ
easat-9471	296	20	rewards	reward	NOUN
easat-9471	296	21	across	across	ADP
easat-9471	296	22	diverse	diverse	ADJ
easat-9471	296	23	domains	domain	NOUN
easat-9471	296	24	,	,	PUNCT
easat-9471	296	25	"	"	PUNCT
easat-9471	296	26	arxiv	arxiv	PROPN
easat-9471	296	27	preprint	preprint	VERB
easat-9471	296	28	arxiv:2503.23829	arxiv:2503.23829	NUM
easat-9471	296	29	,	,	PUNCT
easat-9471	296	30	2025	2025	NUM
easat-9471	296	31	.	.	PUNCT
easat-9471	297	1	[	[	X
easat-9471	297	2	26	26	NUM
easat-9471	297	3	]	]	PUNCT
easat-9471	297	4	j.	j.	PROPN
easat-9471	297	5	wu	wu	PROPN
easat-9471	297	6	,	,	PUNCT
easat-9471	297	7	s.	s.	PROPN
easat-9471	297	8	yin	yin	PROPN
easat-9471	297	9	,	,	PUNCT
easat-9471	297	10	n.	n.	PROPN
easat-9471	297	11	feng	feng	PROPN
easat-9471	297	12	,	,	PUNCT
easat-9471	297	13	and	and	CCONJ
easat-9471	297	14	m.	m.	NOUN
easat-9471	297	15	long	long	ADV
easat-9471	297	16	,	,	PUNCT
easat-9471	297	17	"	"	PUNCT
easat-9471	297	18	rlvr	rlvr	NOUN
easat-9471	297	19	-	-	PUNCT
easat-9471	297	20	world	world	NOUN
easat-9471	297	21	:	:	PUNCT
easat-9471	297	22	training	train	VERB
easat-9471	297	23	world	world	NOUN
easat-9471	297	24	models	model	NOUN
easat-9471	297	25	with	with	ADP
easat-9471	297	26	reinforcement	reinforcement	NOUN
easat-9471	297	27	learning	learning	NOUN
easat-9471	297	28	,	,	PUNCT
easat-9471	297	29	"	"	PUNCT
easat-9471	297	30	arxiv	arxiv	PROPN
easat-9471	297	31	preprint	preprint	NOUN
easat-9471	297	32	arxiv:2505.13934	arxiv:2505.13934	NOUN
easat-9471	297	33	,	,	PUNCT
easat-9471	297	34	2025	2025	NUM
easat-9471	297	35	.	.	PUNCT
easat-9471	298	1	[	[	X
easat-9471	298	2	27	27	NUM
easat-9471	298	3	]	]	PUNCT
easat-9471	298	4	x.	x.	NOUN
easat-9471	298	5	xin	xin	PROPN
easat-9471	298	6	,	,	PUNCT
easat-9471	298	7	t.	t.	PROPN
easat-9471	298	8	pimentel	pimentel	PROPN
easat-9471	298	9	,	,	PUNCT
easat-9471	298	10	a.	a.	NOUN
easat-9471	298	11	karatzoglou	karatzoglou	PROPN
easat-9471	298	12	,	,	PUNCT
easat-9471	298	13	p.	p.	PROPN
easat-9471	298	14	ren	ren	PROPN
easat-9471	298	15	,	,	PUNCT
easat-9471	298	16	k.	k.	PROPN
easat-9471	298	17	christakopoulou	christakopoulou	PROPN
easat-9471	298	18	,	,	PUNCT
easat-9471	298	19	and	and	CCONJ
easat-9471	298	20	z.	z.	PROPN
easat-9471	298	21	ren	ren	PROPN
easat-9471	298	22	,	,	PUNCT
easat-9471	298	23	"	"	PUNCT
easat-9471	298	24	rethinking	rethink	VERB
easat-9471	298	25	reinforcement	reinforcement	NOUN
easat-9471	298	26	learning	learning	NOUN
easat-9471	298	27	for	for	ADP
easat-9471	298	28	recommendation	recommendation	NOUN
easat-9471	298	29	:	:	PUNCT
easat-9471	298	30	a	a	DET
easat-9471	298	31	prompt	prompt	ADJ
easat-9471	298	32	perspective	perspective	NOUN
easat-9471	298	33	,	,	PUNCT
easat-9471	298	34	"	"	PUNCT
easat-9471	298	35	in	in	ADP
easat-9471	298	36	proceedings	proceeding	NOUN
easat-9471	298	37	of	of	ADP
easat-9471	298	38	the	the	DET
easat-9471	298	39	45th	45th	ADJ
easat-9471	298	40	international	international	PROPN
easat-9471	298	41	acm	acm	PROPN
easat-9471	298	42	sigir	sigir	PROPN
easat-9471	298	43	conference	conference	PROPN
easat-9471	298	44	on	on	ADP
easat-9471	298	45	research	research	NOUN
easat-9471	298	46	and	and	CCONJ
easat-9471	298	47	development	development	NOUN
easat-9471	298	48	in	in	ADP
easat-9471	298	49	information	information	NOUN
easat-9471	298	50	retrieval	retrieval	NOUN
easat-9471	298	51	,	,	PUNCT
easat-9471	298	52	2022	2022	NUM
easat-9471	298	53	,	,	PUNCT
easat-9471	298	54	pp	pp	ADJ
easat-9471	298	55	.	.	PUNCT
easat-9471	299	1	1347	1347	NUM
easat-9471	299	2	-	-	SYM
easat-9471	299	3	1357	1357	NUM
easat-9471	299	4	.	.	PUNCT
easat-9471	300	1	881	881	NUM
easat-9471	300	2	edelweiss	edelweiss	PROPN
easat-9471	300	3	applied	apply	VERB
easat-9471	300	4	science	science	NOUN
easat-9471	300	5	and	and	CCONJ
easat-9471	300	6	technology	technology	NOUN
easat-9471	300	7	issn	issn	PROPN
easat-9471	300	8	:	:	PUNCT
easat-9471	300	9	2576	2576	NUM
easat-9471	300	10	-	-	SYM
easat-9471	300	11	8484	8484	NUM
easat-9471	300	12	vol	vol	NOUN
easat-9471	300	13	.	.	PROPN
easat-9471	301	1	9	9	NUM
easat-9471	301	2	,	,	PUNCT
easat-9471	301	3	no	no	INTJ
easat-9471	301	4	.	.	NOUN
easat-9471	301	5	8	8	NUM
easat-9471	301	6	:	:	PUNCT
easat-9471	301	7	871	871	NUM
easat-9471	301	8	-	-	SYM
easat-9471	301	9	881	881	NUM
easat-9471	301	10	,	,	PUNCT
easat-9471	301	11	2025	2025	NUM
easat-9471	301	12	doi	doi	NOUN
easat-9471	301	13	:	:	PUNCT
easat-9471	301	14	10.55214/2576	10.55214/2576	NUM
easat-9471	301	15	-	-	PUNCT
easat-9471	301	16	8484.v9i8.9471	8484.v9i8.9471	NOUN
easat-9471	302	1	©	©	PROPN
easat-9471	302	2	2025	2025	NUM
easat-9471	302	3	by	by	ADP
easat-9471	302	4	the	the	DET
easat-9471	302	5	authors	author	NOUN
easat-9471	302	6	;	;	PUNCT
easat-9471	302	7	licensee	licensee	PROPN
easat-9471	302	8	learning	learning	NOUN
easat-9471	302	9	gate	gate	NOUN
easat-9471	303	1	[	[	X
easat-9471	303	2	28	28	NUM
easat-9471	303	3	]	]	PUNCT
easat-9471	303	4	z.	z.	PROPN
easat-9471	303	5	ren	ren	PROPN
easat-9471	303	6	et	et	PROPN
easat-9471	303	7	al	al	PROPN
easat-9471	303	8	.	.	PROPN
easat-9471	303	9	,	,	PUNCT
easat-9471	303	10	"	"	PUNCT
easat-9471	303	11	contrastive	contrastive	ADJ
easat-9471	303	12	state	state	NOUN
easat-9471	303	13	augmentations	augmentation	NOUN
easat-9471	303	14	for	for	ADP
easat-9471	303	15	reinforcement	reinforcement	NOUN
easat-9471	303	16	learning	learning	NOUN
easat-9471	303	17	-	-	PUNCT
easat-9471	303	18	based	base	VERB
easat-9471	303	19	recommender	recommender	NOUN
easat-9471	303	20	systems	system	NOUN
easat-9471	303	21	,	,	PUNCT
easat-9471	303	22	"	"	PUNCT
easat-9471	303	23	in	in	ADP
easat-9471	303	24	proceedings	proceeding	NOUN
easat-9471	303	25	of	of	ADP
easat-9471	303	26	the	the	DET
easat-9471	303	27	46th	46th	ADJ
easat-9471	303	28	international	international	PROPN
easat-9471	303	29	acm	acm	PROPN
easat-9471	303	30	sigir	sigir	PROPN
easat-9471	303	31	conference	conference	PROPN
easat-9471	303	32	on	on	ADP
easat-9471	303	33	research	research	NOUN
easat-9471	303	34	and	and	CCONJ
easat-9471	303	35	development	development	NOUN
easat-9471	303	36	in	in	ADP
easat-9471	303	37	information	information	NOUN
easat-9471	303	38	retrieval	retrieval	NOUN
easat-9471	303	39	,	,	PUNCT
easat-9471	303	40	2023	2023	NUM
easat-9471	303	41	,	,	PUNCT
easat-9471	303	42	pp	pp	ADJ
easat-9471	303	43	.	.	PUNCT
easat-9471	304	1	922	922	NUM
easat-9471	304	2	-	-	NUM
easat-9471	304	3	931	931	NUM
easat-9471	304	4	.	.	PUNCT
easat-9471	305	1	[	[	X
easat-9471	305	2	29	29	NUM
easat-9471	305	3	]	]	X
easat-9471	305	4	c.	c.	PROPN
easat-9471	305	5	li	li	PROPN
easat-9471	305	6	et	et	PROPN
easat-9471	305	7	al	al	PROPN
easat-9471	305	8	.	.	PROPN
easat-9471	305	9	,	,	PUNCT
easat-9471	305	10	"	"	PUNCT
easat-9471	305	11	model	model	NOUN
easat-9471	305	12	-	-	PUNCT
easat-9471	305	13	enhanced	enhance	VERB
easat-9471	305	14	contrastive	contrastive	ADJ
easat-9471	305	15	reinforcement	reinforcement	NOUN
easat-9471	305	16	learning	learning	NOUN
easat-9471	305	17	for	for	ADP
easat-9471	305	18	sequential	sequential	ADJ
easat-9471	305	19	recommendation	recommendation	NOUN
easat-9471	305	20	,	,	PUNCT
easat-9471	305	21	"	"	PUNCT
easat-9471	305	22	arxiv	arxiv	PROPN
easat-9471	305	23	preprint	preprint	NOUN
easat-9471	305	24	arxiv:2310.16566	arxiv:2310.16566	NOUN
easat-9471	305	25	,	,	PUNCT
easat-9471	305	26	2023	2023	NUM
easat-9471	305	27	.	.	PUNCT
easat-9471	306	1	[	[	X
easat-9471	306	2	30	30	NUM
easat-9471	306	3	]	]	X
easat-9471	306	4	w.	w.	PROPN
easat-9471	306	5	hamilton	hamilton	PROPN
easat-9471	306	6	,	,	PUNCT
easat-9471	306	7	z.	z.	PROPN
easat-9471	306	8	ying	ying	PROPN
easat-9471	306	9	,	,	PUNCT
easat-9471	306	10	and	and	CCONJ
easat-9471	306	11	j.	j.	PROPN
easat-9471	306	12	leskovec	leskovec	PROPN
easat-9471	306	13	,	,	PUNCT
easat-9471	306	14	"	"	PUNCT
easat-9471	306	15	inductive	inductive	ADJ
easat-9471	306	16	representation	representation	NOUN
easat-9471	306	17	learning	learn	VERB
easat-9471	306	18	on	on	ADP
easat-9471	306	19	large	large	ADJ
easat-9471	306	20	graphs	graph	NOUN
easat-9471	306	21	,	,	PUNCT
easat-9471	306	22	"	"	PUNCT
easat-9471	306	23	advances	advance	NOUN
easat-9471	306	24	in	in	ADP
easat-9471	306	25	neural	neural	ADJ
easat-9471	306	26	information	information	NOUN
easat-9471	306	27	processing	processing	NOUN
easat-9471	306	28	systems	system	NOUN
easat-9471	306	29	,	,	PUNCT
easat-9471	306	30	vol	vol	NOUN
easat-9471	306	31	.	.	PROPN
easat-9471	306	32	30	30	NUM
easat-9471	306	33	,	,	PUNCT
easat-9471	306	34	pp	pp	ADJ
easat-9471	306	35	.	.	PUNCT
easat-9471	306	36	1024–1034	1024–1034	NUM
easat-9471	306	37	,	,	PUNCT
easat-9471	306	38	2017	2017	NUM
easat-9471	306	39	.	.	PUNCT
easat-9471	307	1	[	[	X
easat-9471	307	2	31	31	NUM
easat-9471	307	3	]	]	X
easat-9471	307	4	y.	y.	PROPN
easat-9471	307	5	mroueh	mroueh	PROPN
easat-9471	307	6	,	,	PUNCT
easat-9471	307	7	"	"	PUNCT
easat-9471	307	8	reinforcement	reinforcement	NOUN
easat-9471	307	9	learning	learning	NOUN
easat-9471	307	10	with	with	ADP
easat-9471	307	11	verifiable	verifiable	ADJ
easat-9471	307	12	rewards	reward	NOUN
easat-9471	307	13	:	:	PUNCT
easat-9471	307	14	grpo	grpo	NOUN
easat-9471	307	15	's	's	PART
easat-9471	307	16	effective	effective	ADJ
easat-9471	307	17	loss	loss	NOUN
easat-9471	307	18	,	,	PUNCT
easat-9471	307	19	dynamics	dynamic	NOUN
easat-9471	307	20	,	,	PUNCT
easat-9471	307	21	and	and	CCONJ
easat-9471	307	22	success	success	NOUN
easat-9471	307	23	amplification	amplification	NOUN
easat-9471	307	24	,	,	PUNCT
easat-9471	307	25	"	"	PUNCT
easat-9471	307	26	arxiv	arxiv	PROPN
easat-9471	307	27	preprint	preprint	NOUN
easat-9471	307	28	arxiv:2503.06639	arxiv:2503.06639	PROPN
easat-9471	307	29	,	,	PUNCT
easat-9471	307	30	2025	2025	NUM
easat-9471	307	31	.	.	PUNCT
easat-9471	308	1	[	[	X
easat-9471	308	2	32	32	NUM
easat-9471	308	3	]	]	X
easat-9471	308	4	y.	y.	PROPN
easat-9471	308	5	hou	hou	PROPN
easat-9471	308	6	,	,	PUNCT
easat-9471	308	7	j.	j.	PROPN
easat-9471	308	8	li	li	PROPN
easat-9471	308	9	,	,	PUNCT
easat-9471	308	10	z.	z.	PROPN
easat-9471	308	11	he	he	PRON
easat-9471	308	12	,	,	PUNCT
easat-9471	308	13	a.	a.	PROPN
easat-9471	308	14	yan	yan	PROPN
easat-9471	308	15	,	,	PUNCT
easat-9471	308	16	x.	x.	PROPN
easat-9471	308	17	chen	chen	PROPN
easat-9471	308	18	,	,	PUNCT
easat-9471	308	19	and	and	CCONJ
easat-9471	308	20	j.	j.	PROPN
easat-9471	308	21	mcauley	mcauley	PROPN
easat-9471	308	22	,	,	PUNCT
easat-9471	308	23	"	"	PUNCT
easat-9471	308	24	bridging	bridge	VERB
easat-9471	308	25	language	language	NOUN
easat-9471	308	26	and	and	CCONJ
easat-9471	308	27	items	item	NOUN
easat-9471	308	28	for	for	ADP
easat-9471	308	29	retrieval	retrieval	NOUN
easat-9471	308	30	and	and	CCONJ
easat-9471	308	31	recommendation	recommendation	NOUN
easat-9471	308	32	,	,	PUNCT
easat-9471	308	33	"	"	PUNCT
easat-9471	308	34	arxiv	arxiv	PROPN
easat-9471	308	35	preprint	preprint	PROPN
easat-9471	308	36	arxiv:2403.03952	arxiv:2403.03952	PROPN
easat-9471	308	37	,	,	PUNCT
easat-9471	308	38	2024	2024	NUM
easat-9471	308	39	.	.	PUNCT
easat-9471	309	1	[	[	X
easat-9471	309	2	33	33	NUM
easat-9471	309	3	]	]	PUNCT
easat-9471	309	4	w.	w.	PROPN
easat-9471	309	5	x.	x.	PROPN
easat-9471	309	6	zhao	zhao	PROPN
easat-9471	309	7	et	et	PROPN
easat-9471	309	8	al	al	PROPN
easat-9471	309	9	.	.	PROPN
easat-9471	309	10	,	,	PUNCT
easat-9471	309	11	"	"	PUNCT
easat-9471	309	12	recbole	recbole	VERB
easat-9471	309	13	:	:	PUNCT
easat-9471	309	14	towards	towards	ADP
easat-9471	309	15	a	a	DET
easat-9471	309	16	unified	unified	ADJ
easat-9471	309	17	,	,	PUNCT
easat-9471	309	18	comprehensive	comprehensive	ADJ
easat-9471	309	19	and	and	CCONJ
easat-9471	309	20	efficient	efficient	ADJ
easat-9471	309	21	framework	framework	NOUN
easat-9471	309	22	for	for	ADP
easat-9471	309	23	recommendation	recommendation	NOUN
easat-9471	309	24	algorithms	algorithm	NOUN
easat-9471	309	25	,	,	PUNCT
easat-9471	309	26	"	"	PUNCT
easat-9471	309	27	in	in	ADP
easat-9471	309	28	proceedings	proceeding	NOUN
easat-9471	309	29	of	of	ADP
easat-9471	309	30	the	the	DET
easat-9471	309	31	30th	30th	ADJ
easat-9471	309	32	acm	acm	PROPN
easat-9471	309	33	international	international	ADJ
easat-9471	309	34	conference	conference	NOUN
easat-9471	309	35	on	on	ADP
easat-9471	309	36	information	information	NOUN
easat-9471	309	37	&	&	CCONJ
easat-9471	309	38	knowledge	knowledge	PROPN
easat-9471	309	39	management	management	PROPN
easat-9471	309	40	,	,	PUNCT
easat-9471	309	41	2021	2021	NUM
easat-9471	309	42	,	,	PUNCT
easat-9471	309	43	pp	pp	ADJ
easat-9471	309	44	.	.	PUNCT
easat-9471	310	1	4653	4653	NUM
easat-9471	310	2	-	-	SYM
easat-9471	310	3	4664	4664	NUM
easat-9471	310	4	.	.	PUNCT
easat-9471	311	1	[	[	X
easat-9471	311	2	34	34	NUM
easat-9471	311	3	]	]	X
easat-9471	311	4	w.	w.	PROPN
easat-9471	311	5	krichene	krichene	PROPN
easat-9471	311	6	and	and	CCONJ
easat-9471	311	7	s.	s.	PROPN
easat-9471	311	8	rendle	rendle	PROPN
easat-9471	311	9	,	,	PUNCT
easat-9471	311	10	"	"	PUNCT
easat-9471	311	11	on	on	ADP
easat-9471	311	12	sampled	sample	VERB
easat-9471	311	13	metrics	metric	NOUN
easat-9471	311	14	for	for	ADP
easat-9471	311	15	item	item	NOUN
easat-9471	311	16	recommendation	recommendation	NOUN
easat-9471	311	17	,	,	PUNCT
easat-9471	311	18	"	"	PUNCT
easat-9471	311	19	in	in	ADP
easat-9471	311	20	proceedings	proceeding	NOUN
easat-9471	311	21	of	of	ADP
easat-9471	311	22	the	the	DET
easat-9471	311	23	26th	26th	ADJ
easat-9471	311	24	acm	acm	NOUN
easat-9471	311	25	sigkdd	sigkdd	NOUN
easat-9471	311	26	international	international	ADJ
easat-9471	311	27	conference	conference	NOUN
easat-9471	311	28	on	on	ADP
easat-9471	311	29	knowledge	knowledge	PROPN
easat-9471	311	30	discovery	discovery	PROPN
easat-9471	311	31	&	&	CCONJ
easat-9471	311	32	data	data	PROPN
easat-9471	311	33	mining	mining	NOUN
easat-9471	311	34	,	,	PUNCT
easat-9471	311	35	2020	2020	NUM
easat-9471	311	36	,	,	PUNCT
easat-9471	311	37	pp	pp	ADP
easat-9471	311	38	.	.	PUNCT
easat-9471	311	39	1748	1748	NUM
easat-9471	311	40	-	-	SYM
easat-9471	311	41	1757	1757	NUM
easat-9471	311	42	.	.	PUNCT
