{"id":"https://openalex.org/W7134230074","doi":"https://doi.org/10.48550/arxiv.2603.06495","title":"COLD-Steer: Steering Large Language Models via In-Context One-step Learning Dynamics","display_name":"COLD-Steer: Steering Large Language Models via In-Context One-step Learning Dynamics","publication_year":2026,"publication_date":"2026-03-06","ids":{"openalex":"https://openalex.org/W7134230074","doi":"https://doi.org/10.48550/arxiv.2603.06495"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2603.06495","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},"type":"article","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":null,"any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5128557451","display_name":"Kartik Sharma","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sharma, Kartik","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5128431803","display_name":"Rakshit S. Trivedi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Trivedi, Rakshit S.","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":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.25917124,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"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":0.3059999942779541,"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":0.3059999942779541,"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":0.15330000221729279,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.07109999656677246,"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/set","display_name":"Set (abstract data type)","score":0.6150000095367432},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6104000210762024},{"id":"https://openalex.org/keywords/variety","display_name":"Variety (cybernetics)","score":0.5756999850273132},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.5170999765396118},{"id":"https://openalex.org/keywords/kernel","display_name":"Kernel (algebra)","score":0.4778999984264374},{"id":"https://openalex.org/keywords/gradient-descent","display_name":"Gradient descent","score":0.46869999170303345},{"id":"https://openalex.org/keywords/control","display_name":"Control (management)","score":0.4595000147819519},{"id":"https://openalex.org/keywords/face","display_name":"Face (sociological concept)","score":0.4490000009536743}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.736299991607666},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.6150000095367432},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6104000210762024},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.5756999850273132},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.5170999765396118},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4878000020980835},{"id":"https://openalex.org/C74193536","wikidata":"https://www.wikidata.org/wiki/Q574844","display_name":"Kernel (algebra)","level":2,"score":0.4778999984264374},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.47690001130104065},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.46869999170303345},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.4595000147819519},{"id":"https://openalex.org/C2779304628","wikidata":"https://www.wikidata.org/wiki/Q3503480","display_name":"Face (sociological concept)","level":2,"score":0.4490000009536743},{"id":"https://openalex.org/C2777472644","wikidata":"https://www.wikidata.org/wiki/Q16968992","display_name":"Approximate inference","level":3,"score":0.41609999537467957},{"id":"https://openalex.org/C145912823","wikidata":"https://www.wikidata.org/wiki/Q113558","display_name":"Dynamics (music)","level":2,"score":0.36570000648498535},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3479999899864197},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.31029999256134033},{"id":"https://openalex.org/C77405623","wikidata":"https://www.wikidata.org/wiki/Q598451","display_name":"System dynamics","level":2,"score":0.29679998755455017},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.2921999990940094},{"id":"https://openalex.org/C196340769","wikidata":"https://www.wikidata.org/wiki/Q7698910","display_name":"Temporal difference learning","level":3,"score":0.28619998693466187},{"id":"https://openalex.org/C134261354","wikidata":"https://www.wikidata.org/wiki/Q938438","display_name":"Statistical inference","level":2,"score":0.2793999910354614},{"id":"https://openalex.org/C148764684","wikidata":"https://www.wikidata.org/wiki/Q621751","display_name":"Approximation algorithm","level":2,"score":0.262800008058548}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2603.06495","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},{"id":"doi:10.48550/arxiv.2603.06495","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.06495","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":"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":"pmh:doi:10.48550/arxiv.2603.06495","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Article"},"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":{"Activation":[0],"steering":[1,25,139,149],"methods":[2,22,31],"enable":[3],"inference-time":[4],"control":[5,190],"of":[6,41,74,80,131,138,202],"large":[7],"language":[8],"model":[9,189],"(LLM)":[10],"behavior":[11],"without":[12,89,167],"retraining,":[13],"but":[14],"current":[15],"approaches":[16],"face":[17],"a":[18,46,77,101,122,136],"fundamental":[19],"trade-off:":[20],"sample-efficient":[21],"suboptimally":[23],"capture":[24],"signals":[26,36],"from":[27,61],"labeled":[28],"examples,":[29,119],"while":[30,151],"that":[32,49,58,71,106,144,191],"better":[33],"extract":[34],"these":[35],"require":[37],"hundreds":[38],"to":[39,115,158],"thousands":[40],"examples.":[42,66],"We":[43,93],"introduce":[44],"COLD-Steer,":[45],"training-free":[47],"framework":[48,182],"steers":[50],"LLM":[51],"activations":[52,109],"by":[53],"approximating":[54],"the":[55,72,108,159],"representational":[56],"changes":[57],"would":[59],"result":[60],"gradient":[62],"descent":[63],"on":[64,76,177],"in-context":[65],"Our":[67,181],"key":[68],"insight":[69],"is":[70],"effect":[73],"fine-tuning":[75],"small":[78],"set":[79],"examples":[81],"can":[82,192],"be":[83],"efficiently":[84],"approximated":[85],"at":[86],"inference":[87],"time":[88],"actual":[90],"parameter":[91],"updates.":[92],"formalize":[94],"this":[95],"through":[96,174,199],"two":[97,127],"complementary":[98],"approaches:":[99],"(i)":[100],"unit":[102],"kernel":[103],"approximation":[104,124,201],"method":[105],"updates":[107],"directly":[110],"using":[111,152],"gradients":[112],"with":[113],"respect":[114],"them,":[116],"normalized":[117],"across":[118,135],"and":[120,141],"(ii)":[121],"finite-difference":[123],"requiring":[125],"only":[126],"forward":[128],"passes":[129],"regardless":[130],"example":[132],"count.":[133],"Experiments":[134],"variety":[137],"tasks":[140],"benchmarks":[142],"demonstrate":[143],"COLD-Steer":[145,162],"achieves":[146],"upto":[147],"95%":[148],"effectiveness":[150],"50":[153],"times":[154],"fewer":[155],"samples":[156],"compared":[157],"best":[160],"baseline.":[161],"facilitates":[163],"accommodating":[164],"diverse":[165],"perspectives":[166],"extensive":[168],"demonstration":[169],"data,":[170],"which":[171],"we":[172],"validate":[173],"our":[175],"experiments":[176],"pluralistic":[178],"alignment":[179],"tasks.":[180],"opens":[183],"new":[184],"possibilities":[185],"for":[186],"adaptive,":[187],"context-aware":[188],"flexibly":[193],"address":[194],"varying":[195],"loss-driven":[196],"human":[197],"preferences":[198],"principled":[200],"learning":[203],"dynamics":[204],"rather":[205],"than":[206],"specialized":[207],"training":[208],"procedures.":[209]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-03-10T00:00:00"}