{"id":"https://openalex.org/W4412945586","doi":"https://doi.org/10.18653/v1/2025.acl-long.393","title":"LLM Braces: Straightening Out LLM Predictions with Relevant Sub-Updates","display_name":"LLM Braces: Straightening Out LLM Predictions with Relevant Sub-Updates","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412945586","doi":"https://doi.org/10.18653/v1/2025.acl-long.393"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.393","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.393","pdf_url":"https://aclanthology.org/2025.acl-long.393.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.acl-long.393.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100747543","display_name":"Ying Shen","orcid":"https://orcid.org/0000-0001-8086-7189"},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Ying Shen","raw_affiliation_strings":["Computer Science Department University of Illinois Urbana-Champaign Computer Science Department UC Davis"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Computer Science Department University of Illinois Urbana-Champaign Computer Science Department UC Davis","institution_ids":["https://openalex.org/I157725225"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5119204548","display_name":"Lifu Huang","orcid":null},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Lifu Huang","raw_affiliation_strings":["Computer Science Department University of Illinois Urbana-Champaign Computer Science Department UC Davis"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Computer Science Department University of Illinois Urbana-Champaign Computer Science Department UC Davis","institution_ids":["https://openalex.org/I157725225"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I157725225"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.16878227,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"7977","last_page":"7992"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T13643","display_name":"Artificial Intelligence in Law","score":0.7372999787330627,"subfield":{"id":"https://openalex.org/subfields/3320","display_name":"Political Science and International Relations"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T13643","display_name":"Artificial Intelligence in Law","score":0.7372999787330627,"subfield":{"id":"https://openalex.org/subfields/3320","display_name":"Political Science and International Relations"},"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/T11986","display_name":"Scientific Computing and Data Management","score":0.6804999709129333,"subfield":{"id":"https://openalex.org/subfields/1802","display_name":"Information Systems and Management"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T14330","display_name":"Library Science and Information Systems","score":0.6779000163078308,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/computer-science","display_name":"Computer science","score":0.5870720148086548}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5870720148086548}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.393","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.393","pdf_url":"https://aclanthology.org/2025.acl-long.393.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.393","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.393","pdf_url":"https://aclanthology.org/2025.acl-long.393.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G6671297155","display_name":null,"funder_award_id":"CAREER","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"}],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412945586.pdf","grobid_xml":"https://content.openalex.org/works/W4412945586.grobid-xml"},"referenced_works_count":3,"referenced_works":["https://openalex.org/W3168867926","https://openalex.org/W4308023630","https://openalex.org/W4387687067"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Recent":[0],"findings":[1],"reveal":[2],"that":[3,47,57,94,167],"much":[4,131],"of":[5,32,53,72,114,157],"the":[6,30,42,70,80,112,121],"knowledge":[7],"in":[8,17,102,172,193],"a":[9,37,89,133],"Transformer-based":[10],"Large":[11],"Language":[12],"Model":[13],"(LLM)":[14],"is":[15],"encoded":[16],"its":[18,200],"feed-forward":[19],"(FFN)":[20],"layers,":[21],"where":[22],"each":[23,34],"FFN":[24,103],"layer":[25],"can":[26,62,140],"be":[27,63,141],"interpreted":[28],"as":[29,151],"summation":[31],"sub-updates,":[33],"corresponding":[35],"to":[36,79,109,125,143,185,189],"weighted":[38],"column":[39],"vector":[40],"from":[41],"FFN's":[43],"value":[44,100],"parameter":[45],"matrix":[46],"often":[48],"encodes":[49],"human-interpretable":[50],"concepts.In":[51],"light":[52],"this,":[54],"we":[55],"hypothesize":[56],"model":[58],"performance":[59],"and":[60,66,86,91,105,128,137,165,175,196],"behaviors":[61],"further":[64],"enhanced":[65],"controlled":[67,204],"by":[68],"modulating":[69],"contributions":[71],"these":[73,107],"sub-updates":[74],"based":[75],"on":[76,161],"their":[77],"relevance":[78,96],"input":[81],"or":[82],"target":[83],"output":[84],"style,":[85],"propose":[87],"LLMBRACES,":[88],"novel":[90],"efficient":[92],"method":[93],"computes":[95],"scores":[97,108],"associated":[98],"with":[99],"vectors":[101],"layers":[104],"leverages":[106],"dynamically":[110],"adjust":[111],"contribution":[113],"sub-updates.By":[115],"optimizing":[116],"sub-update":[117],"contributions,":[118],"LLMBRACES":[119,139,191],"refines":[120],"prediction":[122],"process,":[123],"leading":[124],"more":[126],"accurate":[127],"reliable":[129],"outputs,":[130],"like":[132],"'brace'":[134],"providing":[135],"support":[136,144],"stability.Moreover,":[138],"extended":[142],"conditional":[145],"control":[146],"over":[147],"generation":[148,195,206],"characteristics,":[149],"such":[150],"sentiment,":[152],"thereby":[153],"offering":[154],"fine-grained":[155],"steering":[156],"LLM":[158],"outputs.Extensive":[159],"experiments":[160],"various":[162],"LLMs-including":[163],"Qwen2.5-1.5B,Llama2-7B,":[164],"Llama3-8B-demonstrate":[166],"LLM-BRACES":[168],"outperforms":[169],"baseline":[170],"approaches":[171],"both":[173],"fine-tuning":[174],"zero-shot":[176],"settings":[177],"while":[178],"requiring":[179],"significantly":[180],"fewer":[181,187],"tunable":[182],"parameters,":[183],"up":[184],"75%":[186],"compared":[188],"LoRA.Furthermore,":[190],"excels":[192],"sentimentcontrolled":[194],"toxicity":[197],"reduction,":[198],"highlighting":[199],"potential":[201],"for":[202],"flexible,":[203],"text":[205],"across":[207],"applications.":[208]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}