{"id":"https://openalex.org/W7162403791","doi":"https://doi.org/10.48550/arxiv.2605.25850","title":"TIAR: Trajectory-Informed Advantage Reweighting for LLM Abstention Learning","display_name":"TIAR: Trajectory-Informed Advantage Reweighting for LLM Abstention Learning","publication_year":2026,"publication_date":"2026-05-25","ids":{"openalex":"https://openalex.org/W7162403791","doi":"https://doi.org/10.48550/arxiv.2605.25850"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.25850","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.25850","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":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.25850","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5125061261","display_name":"Muyu Pan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pan, Muyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137024042","display_name":"Shu Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Shu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137009172","display_name":"Nan Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Nan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137077813","display_name":"Philip Shin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shin, Philip","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137031994","display_name":"Varun Parekh","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Parekh, Varun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101919131","display_name":"Vijaykrishnan Narayanan","orcid":"https://orcid.org/0000-0001-6266-6068"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Narayanan, Vijaykrishnan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137080157","display_name":"Rui Zhang","orcid":"https://orcid.org/0000-0002-7235-505X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Rui","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":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.39629998803138733,"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":0.39629998803138733,"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/T11147","display_name":"Misinformation and Its Impacts","score":0.07029999792575836,"subfield":{"id":"https://openalex.org/subfields/3312","display_name":"Sociology and Political Science"},"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/T12268","display_name":"Deception detection and forensic psychology","score":0.054499998688697815,"subfield":{"id":"https://openalex.org/subfields/3207","display_name":"Social Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/novelty","display_name":"Novelty","score":0.8288999795913696},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.8072999715805054},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.7390999794006348},{"id":"https://openalex.org/keywords/field","display_name":"Field (mathematics)","score":0.4198000133037567},{"id":"https://openalex.org/keywords/novelty-detection","display_name":"Novelty detection","score":0.36579999327659607},{"id":"https://openalex.org/keywords/empirical-research","display_name":"Empirical research","score":0.36059999465942383}],"concepts":[{"id":"https://openalex.org/C2778738651","wikidata":"https://www.wikidata.org/wiki/Q16546687","display_name":"Novelty","level":2,"score":0.8288999795913696},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.8072999715805054},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.7390999794006348},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7106999754905701},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5764999985694885},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5242999792098999},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.4198000133037567},{"id":"https://openalex.org/C2778924833","wikidata":"https://www.wikidata.org/wiki/Q7064603","display_name":"Novelty detection","level":3,"score":0.36579999327659607},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.36059999465942383},{"id":"https://openalex.org/C64452783","wikidata":"https://www.wikidata.org/wiki/Q1524945","display_name":"Ternary operation","level":2,"score":0.32440000772476196},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.3034999966621399},{"id":"https://openalex.org/C2779439875","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Natural language understanding","level":3,"score":0.2606000006198883},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.25949999690055847},{"id":"https://openalex.org/C2781311116","wikidata":"https://www.wikidata.org/wiki/Q83306","display_name":"Group (periodic table)","level":2,"score":0.25529998540878296},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.25130000710487366}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.25850","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.25850","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:10.48550/arxiv.2605.25850","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.25850","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":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/9","display_name":"Industry, innovation and infrastructure","score":0.43358781933784485}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"This":[0,20],"paper":[1,21,72],"investigates":[2],"large":[3,17],"language":[4,18],"model":[5],"(LLM)":[6],"abstention":[7,39,55,89,130,149,171],"learning,":[8],"specifically":[9],"using":[10],"ternary":[11,29,183],"reward,":[12],"which":[13],"incentivize":[14],"truthfulness":[15],"in":[16,74],"models.":[19],"extends":[22],"that":[23,106,167],"idea":[24],"by":[25],"moving":[26],"from":[27],"a":[28,32,87,94,112],"reward":[30,40,95],"to":[31,97,119,126,143,145],"Trajectory-Informed":[33],"advantage":[34,77],"reweighting,":[35],"dynamically":[36,127],"re-weights":[37],"the":[38,116,120,129,136,146,154,181],"during":[41],"Group":[42],"Relative":[43],"Policy":[44],"Optimization":[45],"(GRPO)":[46],"training.":[47],"The":[48,68],"objective":[49],"of":[50,58,70,115,148,176,187],"this":[51,71,91,140,159],"work":[52,141],"focuses":[53],"on":[54,153,185],"learning":[56],"instead":[57],"improving":[59],"truthfulness,":[60],"serving":[61],"as":[62,86,111,135,139],"an":[63],"exploration":[64],"into":[65],"hallucination":[66],"reduction.":[67],"novelty":[69],"lies":[73],"methodological":[75],"innovation,":[76],"re-weighting,":[78],"and":[79,101,161],"benchmark":[80,155,189],"selection.":[81],"Leveraging":[82],"GRPO's":[83],"multiple":[84],"trajectories":[85,107],"natural":[88],"signal,":[90],"method":[92,160],"uses":[93],"signal":[96],"explore":[98],"knowledge":[99],"boundaries":[100],"encourage":[102],"consistency.":[103],"By":[104],"demonstrating":[105],"can":[108],"be":[109],"used":[110,125,134],"confidence":[113],"indicator":[114],"policy":[117],"relative":[118],"query,":[121],"they":[122],"are":[123],"then":[124],"calculate":[128],"advantage.":[131],"AbstentionBench":[132],"is":[133],"evaluation":[137,178],"benchmark,":[138],"aims":[142],"contribute":[144],"field":[147],"learning.":[150],"All":[151],"datasets":[152,190],"were":[156],"tested":[157],"against":[158],"various":[162],"baselines.":[163],"Empirical":[164],"results":[165],"demonstrate":[166],"TIAR":[168],"achieves":[169],"state-of-the-art":[170],"F1":[172],"scores":[173],"across":[174],"five":[175],"six":[177],"categories,":[179],"outperforming":[180],"static":[182],"baseline":[184,194],"17":[186],"31":[188],"while":[191],"fully":[192],"preserving":[193],"accuracy.":[195]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-05-27T00:00:00"}