{"id":"https://openalex.org/W7224251169","doi":"https://doi.org/10.18653/2026.findings-emnlp.934","title":"Skip-Connected Rollouts for Mid-Trajectory Rewards in GRPO","display_name":"Skip-Connected Rollouts for Mid-Trajectory Rewards in GRPO","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7224251169","doi":"https://doi.org/10.18653/2026.findings-emnlp.934"},"language":"en","primary_location":{"id":"doi:10.18653/2026.findings-emnlp.934","doi":"https://doi.org/10.18653/2026.findings-emnlp.934","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.findings-emnlp.934","pdf_url":"https://aclanthology.org/2026.findings-emnlp.934.pdf","source":{"id":"https://openalex.org/S7407096085","display_name":"Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2026","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","is_traditional":true,"indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"bronze","oa_url":"https://aclanthology.org/2026.findings-emnlp.934.pdf","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5111362587","display_name":"Fengwei Teng","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fengwei Teng","raw_affiliation_strings":["University of Artificial Intelligence"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Artificial Intelligence","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100429463","display_name":"Jinyi Bai","orcid":"https://orcid.org/0000-0002-4136-5730","observed_orcids":["https://orcid.org/0000-0002-4136-5730"]},"institutions":[{"id":"https://openalex.org/I78988378","display_name":"Renmin University of China","ror":"https://ror.org/041pakw92","country_code":"CN","type":"education","lineage":["https://openalex.org/I78988378"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jinyi Bai","raw_affiliation_strings":["Renmin University of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Renmin University of China","institution_ids":["https://openalex.org/I78988378"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108323199","display_name":"Xinhao Yao","orcid":"https://orcid.org/0009-0004-3460-7262","observed_orcids":["https://orcid.org/0009-0004-3460-7262"]},"institutions":[{"id":"https://openalex.org/I78988378","display_name":"Renmin University of China","ror":"https://ror.org/041pakw92","country_code":"CN","type":"education","lineage":["https://openalex.org/I78988378"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Xinhao Yao","raw_affiliation_strings":["Renmin University of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Renmin University of China","institution_ids":["https://openalex.org/I78988378"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139161309","display_name":"Demi Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I74973139","display_name":"Carnegie Mellon University","ror":"https://ror.org/05x2bcf33","country_code":"US","type":"education","lineage":["https://openalex.org/I74973139"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Demi Wang","raw_affiliation_strings":["Carnegie Mellon University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Carnegie Mellon University","institution_ids":["https://openalex.org/I74973139"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5160077641","display_name":"Jiahao Zhao","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I78988378","display_name":"Renmin University of China","ror":"https://ror.org/041pakw92","country_code":"CN","type":"education","lineage":["https://openalex.org/I78988378"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jiahao Zhao","raw_affiliation_strings":["Renmin University of China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Renmin University of China","institution_ids":["https://openalex.org/I78988378"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5160079103","display_name":"Zhijiang Guo","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4405261251","display_name":"The Hong Kong University of Science and Technology (Guangzhou)","ror":"https://ror.org/050h0vm43","country_code":"CN","type":"education","lineage":["https://openalex.org/I4405261251"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhijiang Guo","raw_affiliation_strings":["The Hong Kong University of Science and Technology (Guangzhou )"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"The Hong Kong University of Science and Technology (Guangzhou )","institution_ids":["https://openalex.org/I4405261251"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.81983778,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"19710","last_page":"19728"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.9340999722480774,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.9340999722480774,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.021299999207258224,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.01889999955892563,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/group-relative-policy-optimization","display_name":"Group Relative Policy Optimization","score":0.6639999747276306},{"id":"https://openalex.org/keywords/reinforcement-learning-from-human-feedback","display_name":"reinforcement learning from human feedback","score":0.6579999923706055},{"id":"https://openalex.org/keywords/reward-modeling","display_name":"reward modeling","score":0.5120000243186951}],"concepts":[{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.44200000166893005},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.3456000089645386},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.33739998936653137},{"id":"https://openalex.org/C2778137410","wikidata":"https://www.wikidata.org/wiki/Q2732820","display_name":"Government (linguistics)","level":2,"score":0.32420000433921814},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.29580000042915344},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.2808000147342682},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.26919999718666077},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.2685000002384186},{"id":"https://openalex.org/C145097563","wikidata":"https://www.wikidata.org/wiki/Q1148747","display_name":"Payment","level":2,"score":0.24789999425411224},{"id":"https://openalex.org/C108170787","wikidata":"https://www.wikidata.org/wiki/Q3951828","display_name":"Agency (philosophy)","level":2,"score":0.24410000443458557}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/2026.findings-emnlp.934","doi":"https://doi.org/10.18653/2026.findings-emnlp.934","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.findings-emnlp.934","pdf_url":"https://aclanthology.org/2026.findings-emnlp.934.pdf","source":{"id":"https://openalex.org/S7407096085","display_name":"Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/2026.findings-emnlp.934","doi":"https://doi.org/10.18653/2026.findings-emnlp.934","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.findings-emnlp.934","pdf_url":"https://aclanthology.org/2026.findings-emnlp.934.pdf","source":{"id":"https://openalex.org/S7407096085","display_name":"Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7224251169.pdf","grobid_xml":"https://content.openalex.org/works/W7224251169.grobid-xml"},"referenced_works_count":52,"referenced_works":["https://openalex.org/W4391631327","https://openalex.org/W4391800152","https://openalex.org/W4403852025","https://openalex.org/W4407174758","https://openalex.org/W4407359594","https://openalex.org/W4414634803","https://openalex.org/W4414790631","https://openalex.org/W4414809844","https://openalex.org/W4414854664","https://openalex.org/W4414898977","https://openalex.org/W4414908695","https://openalex.org/W4414975272","https://openalex.org/W4415025740","https://openalex.org/W4415109214","https://openalex.org/W4415136889","https://openalex.org/W4415249932","https://openalex.org/W4415277433","https://openalex.org/W4415317764","https://openalex.org/W4415334799","https://openalex.org/W4415868133","https://openalex.org/W4416133086","https://openalex.org/W4416184383","https://openalex.org/W4416188173","https://openalex.org/W4416256207","https://openalex.org/W4416330276","https://openalex.org/W4416353973","https://openalex.org/W4416373167","https://openalex.org/W4416710896","https://openalex.org/W4416774044","https://openalex.org/W4416937012","https://openalex.org/W4417092538","https://openalex.org/W4417129463","https://openalex.org/W4417351618","https://openalex.org/W4417530951","https://openalex.org/W6884962196","https://openalex.org/W6929246819","https://openalex.org/W6929328961","https://openalex.org/W6929370958","https://openalex.org/W6929629322","https://openalex.org/W6947802607","https://openalex.org/W7077049178","https://openalex.org/W7077055105","https://openalex.org/W7080137253","https://openalex.org/W7083832114","https://openalex.org/W7101782290","https://openalex.org/W7123695520","https://openalex.org/W7123730413","https://openalex.org/W7124424566","https://openalex.org/W7162112465","https://openalex.org/W7166806848","https://openalex.org/W7166807237","https://openalex.org/W7166853295"],"related_works":["https://openalex.org/W7103882922","https://openalex.org/W7160618840","https://openalex.org/W6983778797","https://openalex.org/W7197041240","https://openalex.org/W3212393417","https://openalex.org/W7133233015","https://openalex.org/W7224436734","https://openalex.org/W2801870343","https://openalex.org/W7161491997","https://openalex.org/W7224674240"],"abstract_inverted_index":{"Group":[0],"Relative":[1],"Policy":[2],"Optimization":[3],"(GRPO)":[4],"and":[5,49,58,85,114,134,143],"its":[6],"variants":[7],"are":[8],"effective":[9],"for":[10,110,149],"reinforcement":[11],"learning":[12],"with":[13,139,145,158],"verifiable":[14],"outcome":[15],"rewards,":[16],"but":[17],"incorporating":[18],"intermediate":[19],"credit":[20],"remains":[21],"difficult.We":[22],"show":[23],"that,":[24],"under":[25,167],"practical":[26],"sampling":[27],"budgets,":[28],"token-level":[29],"Monte":[30,118],"Carlo":[31,119],"estimates":[32,52],"at":[33],"early":[34],"reasoning":[35],"positions":[36],"have":[37],"high":[38],"variance;":[39],"group-relative":[40,112],"normalization":[41],"turns":[42],"this":[43],"noise":[44],"into":[45],"frequent":[46],"advantage-sign":[47],"errors":[48],"wrongdirection":[50],"gradients.Later":[51],"instead":[53],"approach":[54],"the":[55,75,97,100,115,130,146],"terminal":[56,108],"reward":[57],"add":[59],"little":[60],"information.We":[61],"propose":[62],"SCRoll,":[63],"a":[64,86,122,126,168],"rollout-side":[65],"module":[66],"that":[67],"adds":[68],"randomized":[69],"single-boundary":[70],"segment":[71,83,98],"rewards":[72],"without":[73],"changing":[74],"downstream":[76,87,104],"optimizer.An":[77],"upstream":[78,116,131],"stream":[79,88],"generates":[80],"an":[81],"early-stopped":[82],"s,":[84],"samples":[89],"G":[90],"continuations":[91],"conditioned":[92],"on":[93],"[s,":[94],"q],":[95],"placing":[96],"before":[99],"original":[101],"problem.The":[102],"shared":[103],"group":[105],"supplies":[106],"both":[107],"outcomes":[109],"any":[111],"objective":[113,128],"segment's":[117],"reward,":[120],"avoiding":[121],"separate":[123],"rollout":[124],"group;":[125],"single-stream":[127],"optimizes":[129],"segment.Across":[132],"Qwen2.5-Math-7B":[133],"Llama-3.2-3B-Instruct,SCRoll":[135],"improves":[136],"most":[137],"pairings":[138],"GRPO,":[140],"GSPO,":[141],"DAPO,":[142],"CISPO,":[144],"clearest":[147],"gains":[148],"CISPO.A":[150],"decoupled":[151],"two-process":[152],"baseline":[153],"reaches":[154],"comparable":[155],"performance":[156],"only":[157],"substantially":[159],"more":[160],"generated":[161],"tokens,":[162],"while":[163],"SCRoll":[164],"performs":[165],"better":[166],"matched":[169],"token":[170],"budget.":[171]},"counts_by_year":[],"updated_date":"2026-11-04T10:54:05.112534","created_date":"2026-10-30T00:00:00"}