{"id":"https://openalex.org/W7166866804","doi":"https://doi.org/10.18653/v1/2026.findings-acl.132","title":"Think Outside the Policy: In-Context Steered Policy Optimization","display_name":"Think Outside the Policy: In-Context Steered Policy Optimization","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166866804","doi":"https://doi.org/10.18653/v1/2026.findings-acl.132"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.132","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.132","pdf_url":"https://aclanthology.org/2026.findings-acl.132.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: ACL 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-acl.132.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139780451","display_name":"Hsiu-Yuan Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hsiu-Yuan Huang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123017886","display_name":"Chenming Tang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chenming Tang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139725186","display_name":"Weijie Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Weijie Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5123014807","display_name":"Clive Bai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Clive Bai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5027918344","display_name":"Saiyong Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Saiyong Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139771503","display_name":"Yunfang Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yunfang Wu","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.9376195,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"2758","last_page":"2776"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12210","display_name":"Policy Transfer and Learning","score":0.07609999924898148,"subfield":{"id":"https://openalex.org/subfields/3320","display_name":"Political Science and International Relations"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T12210","display_name":"Policy Transfer and Learning","score":0.07609999924898148,"subfield":{"id":"https://openalex.org/subfields/3320","display_name":"Political Science and International Relations"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T12617","display_name":"Energy, Environment, and Transportation Policies","score":0.019200000911951065,"subfield":{"id":"https://openalex.org/subfields/2105","display_name":"Renewable Energy, Sustainability and the Environment"},"field":{"id":"https://openalex.org/fields/21","display_name":"Energy"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.01600000075995922,"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/government","display_name":"Government (linguistics)","score":0.32710000872612},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.289900004863739},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.2896000146865845},{"id":"https://openalex.org/keywords/work","display_name":"Work (physics)","score":0.28439998626708984},{"id":"https://openalex.org/keywords/public-policy","display_name":"Public policy","score":0.28029999136924744}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.45559999346733093},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.3831000030040741},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.37229999899864197},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.3474000096321106},{"id":"https://openalex.org/C2778137410","wikidata":"https://www.wikidata.org/wiki/Q2732820","display_name":"Government (linguistics)","level":2,"score":0.32710000872612},{"id":"https://openalex.org/C539667460","wikidata":"https://www.wikidata.org/wiki/Q2414942","display_name":"Management science","level":1,"score":0.2939000129699707},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.289900004863739},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2896000146865845},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.28439998626708984},{"id":"https://openalex.org/C109986646","wikidata":"https://www.wikidata.org/wiki/Q546113","display_name":"Public policy","level":2,"score":0.28029999136924744},{"id":"https://openalex.org/C182306322","wikidata":"https://www.wikidata.org/wiki/Q1779371","display_name":"Order (exchange)","level":2,"score":0.25540000200271606}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.132","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.132","pdf_url":"https://aclanthology.org/2026.findings-acl.132.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: ACL 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.132","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.132","pdf_url":"https://aclanthology.org/2026.findings-acl.132.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: ACL 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166866804.pdf","grobid_xml":"https://content.openalex.org/works/W7166866804.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Existing":[0],"Reinforcement":[1],"Learning":[2],"from":[3,62],"Verifiable":[4],"Rewards":[5],"(RLVR)":[6],"methods,":[7],"such":[8,73],"as":[9],"Group":[10],"Relative":[11],"Policy":[12,86],"Optimization":[13,87],"(GRPO),":[14],"have":[15],"achieved":[16],"remarkable":[17],"progress":[18],"in":[19,48],"improving":[20],"the":[21,43,94,119],"reasoning":[22,168],"capabilities":[23],"of":[24,99],"Large":[25],"Reasoning":[26],"Models":[27],"(LRMs).However,":[28],"they":[29],"exhibit":[30],"limited":[31],"exploration":[32,117],"due":[33],"to":[34,42,54,101,137,147],"reliance":[35,68],"on":[36,166],"on-policy":[37],"rollouts":[38],"which":[39,115],"are":[40,76],"confined":[41],"current":[44,120],"policy's":[45],"distribution,":[46],"resulting":[47],"narrow":[49],"trajectory":[50],"diversity.Recent":[51],"approaches":[52],"attempt":[53],"expand":[55],"policy":[56,121],"coverage":[57],"by":[58],"incorporating":[59],"trajectories":[60,141],"generated":[61],"stronger":[63],"expert":[64,103,113,133,150],"models,":[65],"yet":[66],"this":[67],"increases":[69],"computational":[70],"cost":[71],"and":[72,142,163,173],"advanced":[74,125],"models":[75],"often":[77],"inaccessible.To":[78],"address":[79],"these":[80],"issues,":[81],"we":[82],"propose":[83],"In-Context":[84],"Steered":[85],"(ICPO),":[88],"a":[89,171],"unified":[90],"framework":[91],"that":[92,157],"leverages":[93],"inherent":[95],"incontext":[96],"learning":[97],"capability":[98],"LRMs":[100],"provide":[102],"guidance":[104,151],"using":[105],"existing":[106],"datasets.ICPO":[107],"introduces":[108],"mixed-policy":[109],"GRPO":[110],"with":[111,152],"implicit":[112],"forcing,":[114],"expands":[116],"beyond":[118],"distribution":[122],"without":[123],"requiring":[124],"LRM":[126],"trajectories.To":[127],"further":[128],"stabilize":[129],"optimization,":[130],"ICPO":[131,158],"integrates":[132],"region":[134],"reject":[135],"sampling":[136],"filter":[138],"unreliable":[139],"off-policy":[140],"annealed":[143],"expertbonus":[144],"reward":[145],"shaping":[146],"balance":[148],"early":[149],"later":[153],"autonomous":[154],"improvement.Results":[155],"demonstrate":[156],"consistently":[159],"enhances":[160],"RLVR":[161,175],"performance":[162],"training":[164],"stability":[165],"mathematical":[167],"benchmarks,":[169],"revealing":[170],"scalable":[172],"effective":[174],"paradigm":[176],"for":[177],"LRMs.Our":[178],"code":[179],"is":[180],"available":[181],"at":[182],"https://github.":[183]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}