{"id":"https://openalex.org/W7163676492","doi":"https://doi.org/10.48550/arxiv.2606.05376","title":"SHALA-LLM: Smartly Handling Ambiguous Labels in Aligning LLMs","display_name":"SHALA-LLM: Smartly Handling Ambiguous Labels in Aligning LLMs","publication_year":2026,"publication_date":"2026-06-03","ids":{"openalex":"https://openalex.org/W7163676492","doi":"https://doi.org/10.48550/arxiv.2606.05376"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.05376","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05376","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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","license_id":"https://openalex.org/licenses/cc-by","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.2606.05376","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137945658","display_name":"Jingyao Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Jingyao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137941766","display_name":"Ashley Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Ashley","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5001463799","display_name":"Keane Ong","orcid":"https://orcid.org/0009-0007-0547-2049"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ong, Keane","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137973081","display_name":"Paul Pu Liang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang, Paul Pu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137958481","display_name":"Rosalind Picard","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Picard, Rosalind","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/T10667","display_name":"Emotion and Mood Recognition","score":0.5202000141143799,"subfield":{"id":"https://openalex.org/subfields/3205","display_name":"Experimental and Cognitive Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T10667","display_name":"Emotion and Mood Recognition","score":0.5202000141143799,"subfield":{"id":"https://openalex.org/subfields/3205","display_name":"Experimental and Cognitive Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T10664","display_name":"Sentiment Analysis and Opinion Mining","score":0.1453000009059906,"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":0.0502999983727932,"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/ambiguity","display_name":"Ambiguity","score":0.9243999719619751},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.641700029373169},{"id":"https://openalex.org/keywords/distribution","display_name":"Distribution (mathematics)","score":0.39959999918937683},{"id":"https://openalex.org/keywords/natural","display_name":"Natural (archaeology)","score":0.39309999346733093}],"concepts":[{"id":"https://openalex.org/C2780522230","wikidata":"https://www.wikidata.org/wiki/Q1140419","display_name":"Ambiguity","level":2,"score":0.9243999719619751},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.641700029373169},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5284000039100647},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5109000205993652},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.39959999918937683},{"id":"https://openalex.org/C2776608160","wikidata":"https://www.wikidata.org/wiki/Q4785462","display_name":"Natural (archaeology)","level":2,"score":0.39309999346733093},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3910999894142151},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3783000111579895},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.295199990272522},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.2768000066280365}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.05376","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05376","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.05376","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05376","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Many":[0],"human-centered":[1],"tasks,":[2],"including":[3,137],"natural":[4],"language":[5],"inference":[6],"(NLI)":[7],"and":[8,20,49,134,140],"emotion":[9],"recognition":[10],"(ER),":[11],"have":[12,62],"multiple":[13],"plausible":[14],"interpretations,":[15],"leading":[16],"to":[17,40,84,116,160,171],"label":[18,149],"ambiguity":[19,37,78],"challenging":[21],"disagreements":[22],"across":[23],"human":[24,55],"annotators.":[25],"As":[26],"LLMs":[27,115],"are":[28],"increasingly":[29],"deployed":[30],"in":[31,46],"real-world":[32],"settings,":[33],"faithfully":[34],"modeling":[35,175],"such":[36],"is":[38],"essential":[39],"identify":[41],"contested":[42],"inputs,":[43],"preserve":[44],"variability":[45],"ambiguous":[47,126],"cases,":[48],"capture":[50],"the":[51,163],"full":[52],"distribution":[53],"of":[54,75],"judgments.":[56],"Yet,":[57],"existing":[58],"LLM":[59],"alignment":[60],"approaches":[61],"predominantly":[63],"assumed":[64],"a":[65,94,111],"single":[66],"correct":[67],"label,":[68],"excluding":[69],"annotator":[70,120,148,176],"disagreement":[71,177],"during":[72,128],"optimization.":[73,129],"Instead":[74],"treating":[76],"this":[77],"as":[79,87],"noise,":[80],"we":[81],"show":[82],"how":[83],"treat":[85],"it":[86,154],"information":[88],"that":[89,143,174],"improves":[90,145,167],"model":[91],"behavior":[92],"through":[93],"new":[95,112],"algorithm":[96],"called":[97],"SMARTLY":[98],"HANDLING":[99],"AMBIGUOUS":[100],"LABELS":[101],"IN":[102],"ALIGNING":[103],"LLMS":[104],"(SHALA-LLM).":[105],"This":[106],"reinforcement":[107],"learning":[108],"framework":[109],"provides":[110],"way":[113],"for":[114],"learn":[117],"directly":[118],"from":[119],"distributions":[121],"while":[122],"dynamically":[123],"prioritizing":[124],"highly":[125],"samples":[127],"Experiments":[130],"on":[131,152],"ambiguity-sensitive":[132],"NLI":[133],"ER":[135],"benchmarks,":[136],"ChaosNLI,":[138,153],"GoEmotions,":[139],"MSP-Podcast,":[141],"demonstrate":[142],"SHALA-LLM":[144,166],"agreement":[146],"with":[147],"distributions,":[150],"e.g.":[151],"reduces":[155],"Jensen-Shannon":[156],"Distance":[157],"by":[158,169],"up":[159,170],"62.1%.":[161],"At":[162],"same":[164],"time,":[165],"F1":[168],"16.7%,":[172],"showing":[173],"can":[178],"also":[179],"strengthen":[180],"classification":[181],"performance.":[182]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-06T00:00:00"}