{"id":"https://openalex.org/W4404132457","doi":"https://doi.org/10.1162/tacl_a_00708","title":"Characterizing Learning Curves During Language Model Pre-Training: Learning, Forgetting, and Stability","display_name":"Characterizing Learning Curves During Language Model Pre-Training: Learning, Forgetting, and Stability","publication_year":2024,"publication_date":"2024-01-01","ids":{"openalex":"https://openalex.org/W4404132457","doi":"https://doi.org/10.1162/tacl_a_00708"},"language":"en","primary_location":{"id":"doi:10.1162/tacl_a_00708","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00708","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00708/2478609/tacl_a_00708.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","doaj"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00708/2478609/tacl_a_00708.pdf","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5046456944","display_name":"Tyler A. Chang","orcid":null},"institutions":[{"id":"https://openalex.org/I36258959","display_name":"University of California San Diego","ror":"https://ror.org/0168r3w48","country_code":"US","type":"education","lineage":["https://openalex.org/I36258959"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Tyler A. Chang","raw_affiliation_strings":["Department of Cognitive Science","Hal\u0131c\u0131o\u011flu Data Science Institute University of California San Diego, USA. tachang@ucsd.edu"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Cognitive Science","institution_ids":[]},{"raw_affiliation_string":"Hal\u0131c\u0131o\u011flu Data Science Institute University of California San Diego, USA. tachang@ucsd.edu","institution_ids":["https://openalex.org/I36258959"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5001760915","display_name":"Zhuowen Tu","orcid":"https://orcid.org/0000-0002-1900-2124"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhuowen Tu","raw_affiliation_strings":["Department of Cognitive Science. ztu@ucsd.edu"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Cognitive Science. ztu@ucsd.edu","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5043344696","display_name":"Benjamin K. Bergen","orcid":"https://orcid.org/0000-0002-9395-9151"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Benjamin K. Bergen","raw_affiliation_strings":["Department of Cognitive Science. bkbergen@ucsd.edu"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Cognitive Science. bkbergen@ucsd.edu","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":2.1871,"has_fulltext":true,"cited_by_count":9,"citation_normalized_percentile":{"value":0.89151489,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":90,"max":99},"biblio":{"volume":"12","issue":null,"first_page":"1346","last_page":"1362"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9998999834060669,"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":0.9998999834060669,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9998000264167786,"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/T12031","display_name":"Speech and dialogue systems","score":0.996399998664856,"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/forgetting","display_name":"Forgetting","score":0.8633227348327637},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.764876127243042},{"id":"https://openalex.org/keywords/stability","display_name":"Stability (learning theory)","score":0.690933346748352},{"id":"https://openalex.org/keywords/learning-curve","display_name":"Learning curve","score":0.6074118614196777},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.5822933912277222},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.5432634353637695},{"id":"https://openalex.org/keywords/language-acquisition","display_name":"Language acquisition","score":0.45649778842926025},{"id":"https://openalex.org/keywords/machine-learning","display_name":"Machine learning","score":0.4344034194946289},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.425920695066452},{"id":"https://openalex.org/keywords/mathematics-education","display_name":"Mathematics education","score":0.20019936561584473},{"id":"https://openalex.org/keywords/cognitive-psychology","display_name":"Cognitive psychology","score":0.1905483603477478},{"id":"https://openalex.org/keywords/psychology","display_name":"Psychology","score":0.1050763726234436}],"concepts":[{"id":"https://openalex.org/C7149132","wikidata":"https://www.wikidata.org/wiki/Q1377840","display_name":"Forgetting","level":2,"score":0.8633227348327637},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.764876127243042},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.690933346748352},{"id":"https://openalex.org/C34585555","wikidata":"https://www.wikidata.org/wiki/Q1368723","display_name":"Learning curve","level":2,"score":0.6074118614196777},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.5822933912277222},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5432634353637695},{"id":"https://openalex.org/C74672266","wikidata":"https://www.wikidata.org/wiki/Q815859","display_name":"Language acquisition","level":2,"score":0.45649778842926025},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4344034194946289},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.425920695066452},{"id":"https://openalex.org/C145420912","wikidata":"https://www.wikidata.org/wiki/Q853077","display_name":"Mathematics education","level":1,"score":0.20019936561584473},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.1905483603477478},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.1050763726234436},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.0},{"id":"https://openalex.org/C111919701","wikidata":"https://www.wikidata.org/wiki/Q9135","display_name":"Operating system","level":1,"score":0.0},{"id":"https://openalex.org/C153294291","wikidata":"https://www.wikidata.org/wiki/Q25261","display_name":"Meteorology","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1162/tacl_a_00708","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00708","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00708/2478609/tacl_a_00708.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"}],"best_oa_location":{"id":"doi:10.1162/tacl_a_00708","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00708","pdf_url":"https://direct.mit.edu/tacl/article-pdf/doi/10.1162/tacl_a_00708/2478609/tacl_a_00708.pdf","source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"sustainable_development_goals":[{"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education","score":0.8500000238418579}],"awards":[{"id":"https://openalex.org/G3714775944","display_name":null,"funder_award_id":"IIS-2127544","funder_id":"https://openalex.org/F4320306076","funder_display_name":"National Science Foundation"},{"id":"https://openalex.org/G6878325487","display_name":"RI: Small: Panoptic 3D Parsing in the Wild","funder_award_id":"2127544","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"},{"id":"https://openalex.org/F4320309480","display_name":"Nvidia","ror":"https://ror.org/03jdj4y14"},{"id":"https://openalex.org/F4320332603","display_name":"University of California, San Diego","ror":"https://ror.org/0168r3w48"},{"id":"https://openalex.org/F4320334227","display_name":"Harvard Data Science Initiative, Harvard University","ror":null}],"has_content":{"pdf":true,"grobid_xml":false},"content_urls":{"pdf":"https://content.openalex.org/works/W4404132457.pdf"},"referenced_works_count":48,"referenced_works":["https://openalex.org/W1951216520","https://openalex.org/W2095689128","https://openalex.org/W2134237567","https://openalex.org/W2164418233","https://openalex.org/W2180912620","https://openalex.org/W2795342569","https://openalex.org/W2795673246","https://openalex.org/W2921890305","https://openalex.org/W2950976310","https://openalex.org/W2979826702","https://openalex.org/W3101284630","https://openalex.org/W3102226577","https://openalex.org/W3103649165","https://openalex.org/W3112717819","https://openalex.org/W3147653952","https://openalex.org/W3173954987","https://openalex.org/W3175560839","https://openalex.org/W3199373335","https://openalex.org/W3200809495","https://openalex.org/W3213448373","https://openalex.org/W4210489638","https://openalex.org/W4212774754","https://openalex.org/W4231510805","https://openalex.org/W4233479333","https://openalex.org/W4287110638","https://openalex.org/W4287890137","https://openalex.org/W4297412003","https://openalex.org/W4309419356","https://openalex.org/W4361766487","https://openalex.org/W4362696539","https://openalex.org/W4385570738","https://openalex.org/W4385571329","https://openalex.org/W4385572950","https://openalex.org/W4385573636","https://openalex.org/W4392240262","https://openalex.org/W4392935324","https://openalex.org/W6639619044","https://openalex.org/W6640820311","https://openalex.org/W6755207826","https://openalex.org/W6761551260","https://openalex.org/W6775935039","https://openalex.org/W6797323639","https://openalex.org/W6799001270","https://openalex.org/W6810220367","https://openalex.org/W6838552513","https://openalex.org/W6843370592","https://openalex.org/W6851579256","https://openalex.org/W7018686672"],"related_works":["https://openalex.org/W4289718052","https://openalex.org/W2164121020","https://openalex.org/W2145559838","https://openalex.org/W2905319430","https://openalex.org/W3116498279","https://openalex.org/W4287549553","https://openalex.org/W4310285384","https://openalex.org/W3183027292","https://openalex.org/W4248896073","https://openalex.org/W2083921152"],"abstract_inverted_index":{"Abstract":[0],"How":[1],"do":[2],"language":[3,22,36,172],"models":[4,37],"learn":[5],"to":[6,44,121],"make":[7],"predictions":[8],"during":[9,124],"pre-training?":[10],"To":[11,72],"study":[12],"this,":[13],"we":[14,77,156,170],"extract":[15],"learning":[16,43,92,163,174,178],"curves":[17,93],"from":[18],"five":[19],"autoregressive":[20],"English":[21],"model":[23,167,173],"pre-training":[24,70,112],"runs,":[25,113],"for":[26,94,158],"1M":[27],"unseen":[28],"tokens":[29,56,96,101],"in":[30,63,97],"context.":[31,98],"We":[32,51],"observe":[33],"that":[34,54,65],"the":[35,79,135,159],"generate":[38,45],"short":[39],"repetitive":[40],"phrases":[41],"before":[42,179],"longer":[46],"and":[47,88,110,117,139,148,169],"more":[48,140,146],"coherent":[49],"text.":[50],"also":[52],"find":[53],"individual":[55,95],"often":[57],"exhibit":[58,106],"sudden":[59],"increases":[60],"or":[61],"decreases":[62],"loss":[64],"are":[66,114,118],"surprisingly":[67],"consistent":[68],"across":[69,111],"runs.":[71],"better":[73],"understand":[74],"these":[75,131],"fluctuations,":[76],"quantify":[78],"final":[80,104],"surprisal,":[81],"within-run":[82],"variability,":[83],"age":[84],"of":[85,91,134,161,182],"acquisition,":[86],"forgettability,":[87],"cross-run":[89],"variability":[90,108],"More":[99],"frequent":[100,141],"reach":[102],"lower":[103],"surprisals,":[105],"less":[107,119],"within":[109],"learned":[115],"earlier,":[116],"likely":[120],"be":[122],"\u201cforgotten\u201d":[123],"pre-training.":[125],"Higher":[126],"n-gram":[127,177,184],"probabilities":[128],"further":[129],"accentuate":[130],"effects.":[132],"Independent":[133],"target":[136],"token,":[137],"shorter":[138],"contexts":[142],"correlate":[143],"with":[144],"marginally":[145],"stable":[147],"quickly":[149],"acquired":[150],"predictions.":[151,185],"Based":[152],"on":[153],"our":[154],"results,":[155],"argue":[157],"existence":[160],"sequential":[162],"dependencies":[164],"between":[165],"different":[166],"capabilities,":[168],"characterize":[171],"as":[175],"early":[176],"gradual":[180],"refinement":[181],"tail":[183]},"counts_by_year":[{"year":2026,"cited_by_count":4},{"year":2025,"cited_by_count":4},{"year":2024,"cited_by_count":1}],"updated_date":"2026-07-23T08:03:31.855105","created_date":"2025-10-10T00:00:00"}