{"id":"https://openalex.org/W7140115172","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.247","title":"SALT: Step-level Advantage Assignment for Long-horizon Agents via Trajectory Graph","display_name":"SALT: Step-level Advantage Assignment for Long-horizon Agents via Trajectory Graph","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7140115172","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.247"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-eacl.247","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.247","pdf_url":"https://aclanthology.org/2026.findings-eacl.247.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EACL 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-eacl.247.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5130324688","display_name":"Jiazheng Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiazheng Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100434922","display_name":"Yawei Wang","orcid":"https://orcid.org/0000-0001-7760-1494"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yawei Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5119858288","display_name":"Qiaojing Yan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qiaojing Yan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130402587","display_name":"Yijun Tian","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yijun Tian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101849784","display_name":"Zhichao Xu","orcid":"https://orcid.org/0009-0002-2475-2031"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhichao Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130402865","display_name":"Huan Song","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huan Song","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130402921","display_name":"Panpan Xu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Panpan Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5130395917","display_name":"Lin Lee Cheong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin Lee Cheong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"4709","last_page":"4725"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11596","display_name":"Constraint Satisfaction and Optimization","score":0.38679999113082886,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T11596","display_name":"Constraint Satisfaction and Optimization","score":0.38679999113082886,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/T11106","display_name":"Data Management and Algorithms","score":0.13369999825954437,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.04490000009536743,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/graph","display_name":"Graph","score":0.43939998745918274},{"id":"https://openalex.org/keywords/trajectory","display_name":"Trajectory","score":0.41370001435279846},{"id":"https://openalex.org/keywords/minification","display_name":"Minification","score":0.3294000029563904},{"id":"https://openalex.org/keywords/path","display_name":"Path (computing)","score":0.28630000352859497},{"id":"https://openalex.org/keywords/matching","display_name":"Matching (statistics)","score":0.28189998865127563},{"id":"https://openalex.org/keywords/graph-theory","display_name":"Graph theory","score":0.27469998598098755}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5511000156402588},{"id":"https://openalex.org/C132525143","wikidata":"https://www.wikidata.org/wiki/Q141488","display_name":"Graph","level":2,"score":0.43939998745918274},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.41370001435279846},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.38029998540878296},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.34369999170303345},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.3294000029563904},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.3084999918937683},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3027999997138977},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.295199990272522},{"id":"https://openalex.org/C2777735758","wikidata":"https://www.wikidata.org/wiki/Q817765","display_name":"Path (computing)","level":2,"score":0.28630000352859497},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.28189998865127563},{"id":"https://openalex.org/C88230418","wikidata":"https://www.wikidata.org/wiki/Q131476","display_name":"Graph theory","level":2,"score":0.27469998598098755},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.2694999873638153},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.26260000467300415},{"id":"https://openalex.org/C41550386","wikidata":"https://www.wikidata.org/wiki/Q529909","display_name":"Multi-agent system","level":2,"score":0.2567000091075897},{"id":"https://openalex.org/C64543145","wikidata":"https://www.wikidata.org/wiki/Q162942","display_name":"Intersection (aeronautics)","level":2,"score":0.2567000091075897},{"id":"https://openalex.org/C146380142","wikidata":"https://www.wikidata.org/wiki/Q1137726","display_name":"Directed graph","level":2,"score":0.25519999861717224}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-eacl.247","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.247","pdf_url":"https://aclanthology.org/2026.findings-eacl.247.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EACL 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-eacl.247","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.247","pdf_url":"https://aclanthology.org/2026.findings-eacl.247.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EACL 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[{"score":0.6867467761039734,"id":"https://metadata.un.org/sdg/13","display_name":"Climate action"}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7140115172.pdf","grobid_xml":"https://content.openalex.org/works/W7140115172.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"have":[4],"demonstrated":[5],"remarkable":[6],"capabilities,":[7],"enabling":[8],"language":[9],"agents":[10],"to":[11,18,79,133,162,194],"excel":[12],"at":[13],"single-turn":[14],"tasks.However,":[15],"their":[16],"application":[17],"complex,":[19],"multi-step,":[20],"and":[21,82,87,104,140,166,176,201],"long-horizon":[22],"tasks":[23],"remains":[24],"challenging.While":[25],"reinforcement":[26],"learning":[27],"(RL)":[28],"offers":[29],"a":[30,46,75,102,109,122,148,191],"promising":[31],"avenue":[32],"for":[33,52],"addressing":[34],"these":[35],"challenges,":[36],"mainstream":[37],"approaches":[38],"typically":[39],"rely":[40],"solely":[41,114],"on":[42,172],"sparse,":[43],"outcome-based":[44],"rewards,":[45],"limitation":[47],"that":[48,107,151,184],"becomes":[49],"especially":[50],"problematic":[51],"group-based":[53,156],"RL":[54,157],"algorithms":[55],"lacking":[56],"critic":[57],"models,":[58],"such":[59,66],"as":[60,147],"Group":[61],"Relative":[62],"Policy":[63],"Optimization":[64],"(GRPO).In":[65],"methods,":[67],"uniformly":[68],"rewarding":[69],"or":[70],"penalizing":[71],"all":[72],"actions":[73,89],"within":[74],"trajectory":[76],"can":[77],"lead":[78],"training":[80],"instability":[81],"suboptimal":[83],"policies,":[84],"because":[85],"beneficial":[86],"detrimental":[88],"are":[90],"often":[91],"entangled":[92],"across":[93],"multi-step":[94],"interactions.To":[95],"address":[96],"this":[97,119],"challenge,":[98],"we":[99],"propose":[100],"SALT,":[101],"novel":[103],"lightweight":[105],"framework":[106],"provides":[108],"finer-grained":[110],"advantage":[111],"assignment,":[112],"derived":[113],"from":[115,124],"outcome":[116],"rewards.We":[117],"achieve":[118],"by":[120],"constructing":[121],"graph":[123],"trajectories":[125],"of":[126,137],"the":[127,135,163,173,196],"same":[128],"prompt,":[129],"which":[130],"allows":[131],"us":[132],"quantify":[134],"quality":[136],"each":[138],"step":[139],"assign":[141],"advantages":[142],"accordingly.Crucially,":[143],"SALT":[144,185,200],"is":[145],"designed":[146],"plug-and-play":[149],"module":[150],"seamlessly":[152],"integrates":[153],"with":[154,179],"existing":[155],"algorithms,":[158],"requiring":[159],"no":[160],"modifications":[161],"rollout":[164],"procedure":[165],"introducing":[167],"negligible":[168],"computational":[169],"overhead.Extensive":[170],"experiments":[171],"WebShop,":[174],"ALFWorld,":[175],"Ap-pWorld":[177],"benchmarks":[178],"various":[180],"model":[181],"sizes":[182],"demonstrate":[183],"consistently":[186],"improves":[187],"performance.We":[188],"also":[189],"conduct":[190],"thorough":[192],"analysis":[193],"validate":[195],"design":[197],"choices":[198],"behind":[199],"offer":[202],"actionable":[203],"insights.":[204]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2026-03-24T00:00:00"}