{"id":"https://openalex.org/W7166660119","doi":"https://doi.org/10.48550/arxiv.2606.29858","title":"Smooth Scaling Laws Hide Stepwise Token Learning","display_name":"Smooth Scaling Laws Hide Stepwise Token Learning","publication_year":2026,"publication_date":"2026-06-29","ids":{"openalex":"https://openalex.org/W7166660119","doi":"https://doi.org/10.48550/arxiv.2606.29858"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.29858","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.29858","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.29858","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5139674758","display_name":"Pingjie Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Pingjie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139665981","display_name":"Zechen Hu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Zechen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139640969","display_name":"Peiru Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Peiru","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124946904","display_name":"Fu Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Fu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139634890","display_name":"Debing Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Debing","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":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/T10181","display_name":"Natural Language Processing Techniques","score":0.24709999561309814,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.24709999561309814,"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/T10028","display_name":"Topic Modeling","score":0.17309999465942383,"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/T12090","display_name":"Language and cultural evolution","score":0.13940000534057617,"subfield":{"id":"https://openalex.org/subfields/3316","display_name":"Cultural Studies"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.792900025844574},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.7549999952316284},{"id":"https://openalex.org/keywords/aggregate","display_name":"Aggregate (composite)","score":0.5716999769210815},{"id":"https://openalex.org/keywords/granularity","display_name":"Granularity","score":0.48089998960494995},{"id":"https://openalex.org/keywords/distribution","display_name":"Distribution (mathematics)","score":0.4771000146865845},{"id":"https://openalex.org/keywords/trajectory","display_name":"Trajectory","score":0.4747999906539917},{"id":"https://openalex.org/keywords/scaling-law","display_name":"Scaling law","score":0.43070000410079956}],"concepts":[{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.792900025844574},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.7549999952316284},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5860999822616577},{"id":"https://openalex.org/C4679612","wikidata":"https://www.wikidata.org/wiki/Q866298","display_name":"Aggregate (composite)","level":2,"score":0.5716999769210815},{"id":"https://openalex.org/C177774035","wikidata":"https://www.wikidata.org/wiki/Q1246948","display_name":"Granularity","level":2,"score":0.48089998960494995},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.4771000146865845},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.4747999906539917},{"id":"https://openalex.org/C2988430800","wikidata":"https://www.wikidata.org/wiki/Q428971","display_name":"Scaling law","level":3,"score":0.43070000410079956},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.42910000681877136},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.4185999929904938},{"id":"https://openalex.org/C156778621","wikidata":"https://www.wikidata.org/wiki/Q1365748","display_name":"Spectrum (functional analysis)","level":2,"score":0.414900004863739},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.38749998807907104},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.3431999981403351},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3027999997138977},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.2833999991416931},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2809999883174896},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2565999925136566},{"id":"https://openalex.org/C98385598","wikidata":"https://www.wikidata.org/wiki/Q1339385","display_name":"Empirical distribution function","level":2,"score":0.25529998540878296},{"id":"https://openalex.org/C121864883","wikidata":"https://www.wikidata.org/wiki/Q677916","display_name":"Statistical physics","level":1,"score":0.25360000133514404}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.29858","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.29858","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":"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":"doi:10.48550/arxiv.2606.29858","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.29858","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"display_name":"Quality Education","score":0.6129978895187378,"id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Language":[0],"model":[1,9],"loss":[2,20,77,136],"follows":[3],"remarkably":[4],"regular":[5],"scaling":[6,64,119,186,209],"laws":[7,65,187],"over":[8],"and":[10,111,124,144,173,198],"data":[11],"size,":[12],"yet":[13],"it":[14],"remains":[15],"unclear":[16],"why":[17],"the":[18,99,128,134,139,152,159,170,192],"aggregate":[19],"should":[21],"exhibit":[22],"a":[23,33,59,94],"power-law":[24],"form.":[25],"Existing":[26],"explanations":[27],"often":[28],"attribute":[29],"this":[30,43,200],"regularity":[31],"to":[32,93,121,163,207,213],"heavy-tailed":[34],"spectrum":[35,96,131],"of":[36,70,194],"pattern":[37],"difficulty":[38],"in":[39,53,88],"natural":[40],"language,":[41],"but":[42,211],"view":[44],"has":[45],"not":[46,205],"been":[47],"directly":[48],"validated":[49],"at":[50],"token-level":[51,60,195],"granularity":[52],"large-scale":[54],"real-data":[55],"training.":[56],"We":[57,148],"present":[58],"framework":[61],"that":[62,83,97,151,185,199],"decomposes":[63],"into":[66],"localized":[67,89],"learning":[68,85,196],"events":[69],"individual":[71],"contextualized":[72],"tokens.":[73],"By":[74],"fitting":[75],"token":[76,84],"trajectories":[78],"with":[79,115],"sigmoids,":[80],"we":[81,168],"show":[82,150],"is":[86,155],"concentrated":[87],"transitions,":[90],"giving":[91],"rise":[92],"learning-time":[95,130],"dominates":[98],"scaling-law":[100],"shape.":[101],"Across":[102],"more":[103],"than":[104],"one":[105],"hundred":[106],"pre-training":[107],"runs":[108],"on":[109],"large":[110],"diverse":[112],"real-language":[113],"corpora":[114],"modern":[116],"LLM":[117],"architectures,":[118],"up":[120],"6B":[122],"parameters":[123],"300B":[125],"training":[126,160,215],"tokens,":[127],"measured":[129],"quantitatively":[132],"reconstructs":[133],"validation":[135],"derivative":[137],"along":[138],"training-step":[140],"$T$,":[141],"data-scale":[142],"$D$,":[143],"model-scale":[145],"$M$":[146],"axes.":[147],"further":[149],"same":[153],"signal":[154],"actionable:":[156],"by":[157,191],"reshaping":[158],"distribution":[161,193,201],"according":[162],"when":[164],"tokens":[165],"become":[166],"learnable,":[167],"alter":[169],"optimization":[171],"trajectory":[172],"achieve":[174],"11\\%":[175],"faster":[176],"validation-loss":[177],"reduction.":[178],"These":[179],"results":[180],"provide":[181],"direct":[182],"empirical":[183],"evidence":[184],"are":[188],"governed":[189],"primarily":[190],"times,":[197],"can":[202],"be":[203],"used":[204],"only":[206],"explain":[208],"behavior":[210],"also":[212],"improve":[214],"performance.":[216]},"counts_by_year":[],"updated_date":"2026-07-01T06:29:00.853634","created_date":"2026-07-01T00:00:00"}