{"id":"https://openalex.org/W4416036256","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.832","title":"Linear-Time Demonstration Selection for In-Context Learning via Gradient Estimation","display_name":"Linear-Time Demonstration Selection for In-Context Learning via Gradient Estimation","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416036256","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.832"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.emnlp-main.832","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.832","pdf_url":"https://aclanthology.org/2025.emnlp-main.832.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.emnlp-main.832.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5012706915","display_name":"Ziniu Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Ziniu Zhang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5026741810","display_name":"Zhenshuo Zhang","orcid":"https://orcid.org/0000-0001-6878-097X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhenshuo Zhang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100683862","display_name":"Dongyue Li","orcid":"https://orcid.org/0009-0001-6187-5367"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dongyue Li","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100364443","display_name":"Lu Wang","orcid":"https://orcid.org/0000-0002-4165-4022"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lu Wang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5042038501","display_name":"Jennifer Dy","orcid":"https://orcid.org/0000-0002-8430-134X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jennifer Dy","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5018959690","display_name":"Hongyang R. Zhang","orcid":"https://orcid.org/0009-0005-9118-8516"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hongyang R. Zhang","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5012706915"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.17495868,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"16470","last_page":"16488"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.44699999690055847,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.44699999690055847,"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/T10057","display_name":"Face and Expression Recognition","score":0.05490000173449516,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.0478999987244606,"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/selection","display_name":"Selection (genetic algorithm)","score":0.5361999869346619},{"id":"https://openalex.org/keywords/feature-selection","display_name":"Feature selection","score":0.37779998779296875},{"id":"https://openalex.org/keywords/estimation","display_name":"Estimation","score":0.35530000925064087},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.33469998836517334},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.328900009393692},{"id":"https://openalex.org/keywords/matching","display_name":"Matching (statistics)","score":0.2721000015735626}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5856000185012817},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.5361999869346619},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.535099983215332},{"id":"https://openalex.org/C148483581","wikidata":"https://www.wikidata.org/wiki/Q446488","display_name":"Feature selection","level":2,"score":0.37779998779296875},{"id":"https://openalex.org/C96250715","wikidata":"https://www.wikidata.org/wiki/Q965330","display_name":"Estimation","level":2,"score":0.35530000925064087},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.33469998836517334},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.328900009393692},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.320499986410141},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.28290000557899475},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2815999984741211},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.2721000015735626},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.2687999904155731},{"id":"https://openalex.org/C153258448","wikidata":"https://www.wikidata.org/wiki/Q1199743","display_name":"Gradient descent","level":3,"score":0.26840001344680786},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.2578999996185303},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.25119999051094055}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.emnlp-main.832","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.832","pdf_url":"https://aclanthology.org/2025.emnlp-main.832.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.emnlp-main.832","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.832","pdf_url":"https://aclanthology.org/2025.emnlp-main.832.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 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G6755165505","display_name":null,"funder_award_id":"award","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G8312550757","display_name":"RI: SMALL: Multitask Learning and Fine-Tuning: Measurements and Algorithms for Representational Transfer","funder_award_id":"2412008","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G848032724","display_name":null,"funder_award_id":"Science","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/W4416036256.pdf","grobid_xml":"https://content.openalex.org/works/W4416036256.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"This":[0],"paper":[1],"introduces":[2],"an":[3,115],"algorithm":[4,140],"to":[5,30,60,102,113,142,181,194,200],"select":[6,25,122],"demonstration":[7],"examples":[8],"for":[9,36,118],"in-context":[10,54],"learning":[11],"of":[12,18,28,67,78,157,168],"a":[13,16,72,92,138],"query":[14],"set.Given":[15],"set":[17,146],"n":[19,29],"examples,":[20],"how":[21],"can":[22],"we":[23,98,107],"quickly":[24],"k":[26,123],"out":[27],"best":[31],"serve":[32],"as":[33],"the":[34,65,79,83,96,109,155,162],"conditioning":[35],"downstream":[37],"inference?This":[38],"problem":[39],"has":[40,58],"broad":[41],"applications":[42],"in":[43,82,137],"prompt":[44],"tuning":[45],"and":[46,121,133,144,152,204],"chain-of-thought":[47],"reasoning.Since":[48],"model":[49,89,131,143],"weights":[50],"remain":[51],"fixed":[52],"during":[53],"learning,":[55],"previous":[56],"work":[57,70],"sought":[59],"design":[61],"methods":[62,208],"based":[63,75,209],"on":[64,76,196,210,215],"similarity":[66],"token":[68],"embeddings.This":[69],"proposes":[71],"new":[73],"approach":[74,87],"gradients":[77,134],"output":[80],"taken":[81],"input":[84,211],"embedding":[85],"space.Our":[86],"estimates":[88],"outputs":[90,132],"through":[91],"first-order":[93],"approximation":[94],"using":[95],"gradients.Then,":[97],"apply":[99],"this":[100],"estimation":[101,164],"multiple":[103],"randomly":[104],"sampled":[105,110],"subsets.Finally,":[106],"aggregate":[108],"subset":[111,184],"outcomes":[112],"form":[114],"influence":[116],"score":[117],"each":[119],"demonstration,":[120],"most":[124],"relevant":[125],"examples.This":[126],"procedure":[127,165],"only":[128],"requires":[129],"precomputing":[130],"once,":[135],"resulting":[136],"linear-time":[139],"relative":[141],"training":[145],"sizes.Extensive":[147],"experiments":[148],"across":[149,176],"various":[150],"models":[151,197],"datasets":[153],"validate":[154],"efficiency":[156],"our":[158],"approach.We":[159],"show":[160],"that":[161,186],"gradient":[163],"yields":[166],"approximations":[167],"full":[169,190],"inference":[170,191],"with":[171,198],"less":[172],"than":[173],"1%":[174],"error":[175],"six":[177],"datasets.This":[178],"allows":[179],"us":[180],"scale":[182],"up":[183,193,199],"selection":[185,207],"would":[187],"otherwise":[188],"run":[189],"by":[192,213],"37.7":[195],"34":[201],"billion":[202],"parameters,":[203],"outperform":[205],"existing":[206],"embeddings":[212],"11%":[214],"average.":[216]},"counts_by_year":[],"updated_date":"2026-04-10T15:06:20.359241","created_date":"2025-11-08T00:00:00"}