{"id":"https://openalex.org/W4417440823","doi":"https://doi.org/10.18653/v1/2026.acl-long.199","title":"Deliberative Searcher: Improving LLM Reliability via Reinforcement Learning with Constraints","display_name":"Deliberative Searcher: Improving LLM Reliability via Reinforcement Learning with Constraints","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W4417440823","doi":"https://doi.org/10.18653/v1/2026.acl-long.199"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2026.acl-long.199","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.199","pdf_url":"https://aclanthology.org/2026.acl-long.199.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":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.acl-long.199.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5021378107","display_name":"Ziang Yin","orcid":"https://orcid.org/0009-0009-5308-1100"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhenyun Yin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100602233","display_name":"Shujie Wang","orcid":"https://orcid.org/0009-0000-3736-2206"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shujie Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5090162303","display_name":"Xuhong Wang","orcid":"https://orcid.org/0000-0002-0944-8667"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xuhong Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078711649","display_name":"Xingjun Ma","orcid":"https://orcid.org/0000-0003-2099-4973"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xingjun Ma","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5022220036","display_name":"Yinchun Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yingchun Wang","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":0.0373158,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"4340","last_page":"4354"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.7272999882698059,"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/T10028","display_name":"Topic Modeling","score":0.7272999882698059,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.06109999865293503,"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"}},{"id":"https://openalex.org/T11273","display_name":"Advanced Graph Neural Networks","score":0.06030000001192093,"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/reinforcement-learning","display_name":"Reinforcement learning","score":0.7734000086784363},{"id":"https://openalex.org/keywords/reliability","display_name":"Reliability (semiconductor)","score":0.7522000074386597},{"id":"https://openalex.org/keywords/trustworthiness","display_name":"Trustworthiness","score":0.5913000106811523},{"id":"https://openalex.org/keywords/calibration","display_name":"Calibration","score":0.5407999753952026},{"id":"https://openalex.org/keywords/reflection","display_name":"Reflection (computer programming)","score":0.4936999976634979},{"id":"https://openalex.org/keywords/certainty","display_name":"Certainty","score":0.48899999260902405},{"id":"https://openalex.org/keywords/component","display_name":"Component (thermodynamics)","score":0.3424000144004822}],"concepts":[{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.7734000086784363},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7670000195503235},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.7522000074386597},{"id":"https://openalex.org/C153701036","wikidata":"https://www.wikidata.org/wiki/Q659974","display_name":"Trustworthiness","level":2,"score":0.5913000106811523},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5691999793052673},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5464000105857849},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.5407999753952026},{"id":"https://openalex.org/C65682993","wikidata":"https://www.wikidata.org/wiki/Q1056451","display_name":"Reflection (computer programming)","level":2,"score":0.4936999976634979},{"id":"https://openalex.org/C7493553","wikidata":"https://www.wikidata.org/wiki/Q1520777","display_name":"Certainty","level":2,"score":0.48899999260902405},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.4172999858856201},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.3424000144004822},{"id":"https://openalex.org/C200601418","wikidata":"https://www.wikidata.org/wiki/Q2193887","display_name":"Reliability engineering","level":1,"score":0.34119999408721924},{"id":"https://openalex.org/C133462117","wikidata":"https://www.wikidata.org/wiki/Q4929239","display_name":"Data collection","level":2,"score":0.33970001339912415},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.32440000772476196},{"id":"https://openalex.org/C201729545","wikidata":"https://www.wikidata.org/wiki/Q2193887","display_name":"Reliability theory","level":3,"score":0.31679999828338623},{"id":"https://openalex.org/C2778067643","wikidata":"https://www.wikidata.org/wiki/Q166507","display_name":"Interval (graph theory)","level":2,"score":0.2793999910354614},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.274399995803833},{"id":"https://openalex.org/C2781002164","wikidata":"https://www.wikidata.org/wiki/Q6822311","display_name":"Meta learning (computer science)","level":3,"score":0.2709999978542328}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2026.acl-long.199","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.199","pdf_url":"https://aclanthology.org/2026.acl-long.199.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":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2507.16727","is_oa":true,"landing_page_url":"https://arxiv.org/abs/2507.16727","pdf_url":"https://arxiv.org/pdf/2507.16727","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by-nc-sa","license_id":"https://openalex.org/licenses/cc-by-nc-sa","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2507.16727","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2507.16727","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.acl-long.199","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.199","pdf_url":"https://aclanthology.org/2026.acl-long.199.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":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G5650502627","display_name":null,"funder_award_id":"62276067","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320329860","display_name":"National Science and Technology Major Project","ror":null}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4417440823.pdf","grobid_xml":"https://content.openalex.org/works/W4417440823.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1,146],"models":[2,77],"with":[3,13,39,75,110],"search":[4,24],"capabilities":[5],"frequently":[6],"exhibit":[7],"miscalibrated":[8],"confidence,":[9],"producing":[10],"incorrect":[11],"answers":[12],"high":[14],"certainty.We":[15],"present":[16],"Deliberative":[17],"Searcher,":[18],"a":[19,114,125],"reasoning-primary":[20],"framework":[21,143],"that":[22],"integrates":[23],"operations":[25],"into":[26],"chain-of-thought":[27],"generation":[28],"while":[29,68],"maintaining":[30],"explicit":[31],"confidence":[32,85,123],"calibration.Our":[33],"method":[34],"employs":[35],"constrained":[36,141],"reinforcement":[37],"learning":[38],"adaptive":[40,132],"Lagrangian":[41],"multipliers":[42],"to":[43,66,82],"jointly":[44],"optimize":[45],"correctness":[46],"and":[47,78,102,131],"reliability.Experiments":[48],"across":[49],"five":[50],"benchmarks":[51],"demonstrate":[52],"substantial":[53],"improvements:":[54],"our":[55,69],"7B":[56],"model":[57],"reduces":[58,79],"average":[59],"false-certain":[60,80],"rates":[61,81],"from":[62],"54%":[63],"in":[64,117,144],"baselines":[65],"2%,":[67],"72B":[70],"variant":[71],"achieves":[72],"competitive":[73],"accuracy":[74],"closedsource":[76],"9%.The":[83],"well-calibrated":[84],"scores":[86],"also":[87],"enable":[88],"more":[89],"efficient":[90],"test-time":[91,133],"compute:":[92],"instead":[93],"of":[94,106,138],"standard":[95],"majority":[96,108],"voting,":[97],"we":[98],"use":[99],"confidence-weighted":[100],"aggregation":[101],"match":[103],"the":[104,136,139],"performance":[105],"16-sample":[107],"voting":[109],"only":[111],"4":[112],"samples,":[113],"4\u00d7":[115],"reduction":[116],"inference":[118],"compute.These":[119],"results":[120],"establish":[121],"calibrated":[122],"as":[124],"foundation":[126],"for":[127],"both":[128],"trustworthy":[129],"outputs":[130],"compute,":[134],"demonstrating":[135],"value":[137],"proposed":[140],"RL":[142],"search-augmented":[145],"models.":[147]},"counts_by_year":[],"updated_date":"2026-08-05T07:39:15.569665","created_date":"2025-10-10T00:00:00"}