{"id":"https://openalex.org/W7154619402","doi":"https://doi.org/10.48550/arxiv.2604.13552","title":"Training-Free Test-Time Contrastive Learning for Large Language Models","display_name":"Training-Free Test-Time Contrastive Learning for Large Language Models","publication_year":2026,"publication_date":"2026-04-15","ids":{"openalex":"https://openalex.org/W7154619402","doi":"https://doi.org/10.48550/arxiv.2604.13552"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.13552","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13552","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.13552","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133767783","display_name":"Kaiwen Zheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Kaiwen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133756309","display_name":"Kai Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Kai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133758352","display_name":"Jinwu Hu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Jinwu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126764161","display_name":"Te Gu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gu, Te","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133730486","display_name":"Mingkai Peng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Peng, Mingkai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133738578","display_name":"Fei Liu","orcid":"https://orcid.org/0009-0004-5499-6357"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Fei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"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":0.603600025177002,"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":0.603600025177002,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.08540000021457672,"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":0.07100000232458115,"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/inference","display_name":"Inference","score":0.6460999846458435},{"id":"https://openalex.org/keywords/adaptation","display_name":"Adaptation (eye)","score":0.44040000438690186},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.4074000120162964},{"id":"https://openalex.org/keywords/code","display_name":"Code (set theory)","score":0.3982999920845032},{"id":"https://openalex.org/keywords/question-answering","display_name":"Question answering","score":0.35260000824928284},{"id":"https://openalex.org/keywords/structured-prediction","display_name":"Structured prediction","score":0.3361000120639801},{"id":"https://openalex.org/keywords/encoding","display_name":"Encoding (memory)","score":0.3346000015735626},{"id":"https://openalex.org/keywords/automated-reasoning","display_name":"Automated reasoning","score":0.32710000872612}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.833899974822998},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6460999846458435},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6403999924659729},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.5781000256538391},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.44040000438690186},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.41839998960494995},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.4074000120162964},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3982999920845032},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.35260000824928284},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.3361000120639801},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.3346000015735626},{"id":"https://openalex.org/C195344581","wikidata":"https://www.wikidata.org/wiki/Q2555318","display_name":"Automated reasoning","level":2,"score":0.32710000872612},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.31850001215934753},{"id":"https://openalex.org/C20162079","wikidata":"https://www.wikidata.org/wiki/Q1151406","display_name":"Case-based reasoning","level":2,"score":0.31790000200271606},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.31610000133514404},{"id":"https://openalex.org/C2164484","wikidata":"https://www.wikidata.org/wiki/Q5170150","display_name":"Core (optical fiber)","level":2,"score":0.3154999911785126},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.26969999074935913},{"id":"https://openalex.org/C155092808","wikidata":"https://www.wikidata.org/wiki/Q182557","display_name":"Computational linguistics","level":2,"score":0.26429998874664307},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.2628999948501587},{"id":"https://openalex.org/C90312973","wikidata":"https://www.wikidata.org/wiki/Q7449052","display_name":"Semantic data model","level":2,"score":0.26269999146461487},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.257099986076355},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.2531000077724457}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.13552","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13552","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.13552","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13552","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"sustainable_development_goals":[{"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4","score":0.5183982849121094}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"demonstrate":[4,157],"strong":[5,162],"reasoning":[6,99,141,151],"capabilities,":[7],"but":[8],"their":[9],"performance":[10],"often":[11],"degrades":[12],"under":[13,169],"distribution":[14],"shift.":[15],"Existing":[16],"test-time":[17],"adaptation":[18,56],"(TTA)":[19],"methods":[20,168],"rely":[21],"on":[22,41,149],"gradient-based":[23],"updates":[24],"that":[25,58,158],"require":[26],"white-box":[27],"access":[28],"and":[29,112,121,153,165],"need":[30],"substantial":[31],"overhead,":[32],"while":[33,143],"training-free":[34,55],"alternatives":[35],"are":[36],"either":[37],"static":[38],"or":[39],"depend":[40],"external":[42],"guidance.":[43],"In":[44],"this":[45],"paper,":[46],"we":[47],"propose":[48],"Training-Free":[49],"Test-Time":[50],"Contrastive":[51,102],"Learning":[52],"TF-TTCL,":[53],"a":[54,60,77],"framework":[57],"enables":[59],"frozen":[61,137],"LLM":[62,138],"to":[63,96,133],"improve":[64],"online":[65,170],"by":[66],"distilling":[67,115],"supervision":[68],"from":[69],"its":[70],"own":[71],"inference":[72,132],"experiences.":[73],"Specifically,":[74],"TF-TTCL":[75,159],"implements":[76],"dynamic":[78],"\"Explore-Reflect-Steer\"":[79],"loop":[80],"through":[81],"three":[82],"core":[83],"modules:":[84],"1)":[85],"Semantic":[86],"Query":[87],"Augmentation":[88],"first":[89],"diversifies":[90],"problem":[91],"views":[92],"via":[93],"multi-agent":[94],"role-playing":[95],"generate":[97],"different":[98],"trajectories;":[100],"2)":[101],"Experience":[103],"Distillation":[104],"then":[105],"captures":[106],"the":[107,136],"semantic":[108],"gap":[109],"between":[110],"superior":[111],"inferior":[113],"trajectories,":[114],"them":[116],"into":[117],"explicit":[118],"textual":[119],"rules;":[120],"3)":[122],"Contextual":[123],"Rule":[124],"Retrieval":[125],"finally":[126],"activates":[127],"these":[128],"stored":[129],"rules":[130],"during":[131],"dynamically":[134],"steer":[135],"toward":[139],"robust":[140],"patterns":[142],"avoiding":[144],"observed":[145],"errors.":[146],"Extensive":[147],"experiments":[148],"closed-ended":[150],"tasks":[152,156],"open-ended":[154],"evaluation":[155],"consistently":[160],"outperforms":[161],"zero-shot":[163],"baselines":[164],"representative":[166],"TTA":[167],"evaluation.":[171],"Code":[172],"is":[173],"available":[174],"at":[175],"https://github.com/KevinSCUTer/TF-TTCL.":[176]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-04-17T00:00:00"}