{"id":"https://openalex.org/W7135026115","doi":"https://doi.org/10.48550/arxiv.2603.10071","title":"Dissecting Chronos: Sparse Autoencoders Reveal Causal Feature Hierarchies in Time Series Foundation Models","display_name":"Dissecting Chronos: Sparse Autoencoders Reveal Causal Feature Hierarchies in Time Series Foundation Models","publication_year":2026,"publication_date":"2026-03-10","ids":{"openalex":"https://openalex.org/W7135026115","doi":"https://doi.org/10.48550/arxiv.2603.10071"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2603.10071","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.10071","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":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.2603.10071","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100703411","display_name":"Anurag Mishra","orcid":"https://orcid.org/0000-0002-4712-3644"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Mishra, Anurag","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5100703411"],"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/T13702","display_name":"Machine Learning in Healthcare","score":0.20669999718666077,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.20669999718666077,"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/T12205","display_name":"Time Series Analysis and Forecasting","score":0.16930000483989716,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T11918","display_name":"Forecasting Techniques and Applications","score":0.1242000013589859,"subfield":{"id":"https://openalex.org/subfields/1803","display_name":"Management Science and Operations Research"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/interpretability","display_name":"Interpretability","score":0.9043999910354614},{"id":"https://openalex.org/keywords/autoencoder","display_name":"Autoencoder","score":0.5817999839782715},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.5716000199317932},{"id":"https://openalex.org/keywords/encode","display_name":"ENCODE","score":0.5260999798774719},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.48100000619888306},{"id":"https://openalex.org/keywords/series","display_name":"Series (stratigraphy)","score":0.45579999685287476},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.4381999969482422}],"concepts":[{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.9043999910354614},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6396999955177307},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6183000206947327},{"id":"https://openalex.org/C101738243","wikidata":"https://www.wikidata.org/wiki/Q786435","display_name":"Autoencoder","level":3,"score":0.5817999839782715},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.5716000199317932},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.5260999798774719},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.48100000619888306},{"id":"https://openalex.org/C143724316","wikidata":"https://www.wikidata.org/wiki/Q312468","display_name":"Series (stratigraphy)","level":2,"score":0.45579999685287476},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.4381999969482422},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.39899998903274536},{"id":"https://openalex.org/C59404180","wikidata":"https://www.wikidata.org/wiki/Q17013334","display_name":"Feature learning","level":2,"score":0.3926999866962433},{"id":"https://openalex.org/C151406439","wikidata":"https://www.wikidata.org/wiki/Q186588","display_name":"Time series","level":2,"score":0.38580000400543213},{"id":"https://openalex.org/C52622490","wikidata":"https://www.wikidata.org/wiki/Q1026626","display_name":"Feature extraction","level":2,"score":0.33480000495910645},{"id":"https://openalex.org/C2780966255","wikidata":"https://www.wikidata.org/wiki/Q5474306","display_name":"Foundation (evidence)","level":2,"score":0.30239999294281006},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.2989000082015991},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.29829999804496765},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.28600001335144043},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.27489998936653137}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2603.10071","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.10071","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":"doi:10.48550/arxiv.2603.10071","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2603.10071","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":false,"raw_source_name":null,"raw_type":"Preprint"},"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":{"Time":[0],"series":[1],"foundation":[2],"models":[3],"(TSFMs)":[4],"are":[5],"increasingly":[6],"deployed":[7],"in":[8,100,110],"high-stakes":[9],"domains,":[10],"yet":[11],"their":[12],"internal":[13],"representations":[14],"remain":[15],"opaque.":[16],"We":[17],"present":[18],"the":[19,73,81,101,111],"first":[20],"application":[21],"of":[22,34,92],"sparse":[23],"autoencoders":[24],"(SAEs)":[25],"to":[26,132],"a":[27,53,63,85],"TSFM,":[28],"training":[29],"TopK":[30],"SAEs":[31],"on":[32,138],"activations":[33],"Chronos-T5-Large":[35],"(710M":[36],"parameters)":[37],"across":[38],"six":[39],"layers.":[40],"Through":[41],"392":[42],"single-feature":[43,104],"ablation":[44,119],"experiments,":[45],"we":[46],"establish":[47],"that":[48,127,135],"every":[49],"ablated":[50],"feature":[51],"produces":[52],"positive":[54],"CRPS":[55,106],"degradation,":[56],"confirming":[57],"causal":[58],"relevance.":[59],"Our":[60],"analysis":[61],"reveals":[62],"depth-dependent":[64],"hierarchy:":[65],"early":[66],"encoder":[67,83,115],"layers":[68],"encode":[69],"low-level":[70],"frequency":[71],"features,":[72,79],"mid-encoder":[74,102],"concentrates":[75],"causally":[76,89],"critical":[77,97],"change-detection":[78],"and":[80,134],"final":[82,114],"compresses":[84],"rich":[86],"but":[87],"less":[88],"important":[90],"taxonomy":[91],"temporal":[93],"concepts.":[94],"The":[95],"most":[96],"features":[98],"reside":[99],"(max":[103],"Delta":[105],"=":[107],"38.61),":[108],"not":[109],"semantically":[112],"richest":[113],"layer,":[116],"where":[117],"progressive":[118],"paradoxically":[120],"improves":[121],"forecast":[122],"quality.":[123],"These":[124],"findings":[125],"demonstrate":[126],"mechanistic":[128],"interpretability":[129],"transfers":[130],"effectively":[131],"TSFMs":[133],"Chronos-T5":[136],"relies":[137],"abrupt-dynamics":[139],"detection":[140],"rather":[141],"than":[142],"periodic":[143],"pattern":[144],"recognition.":[145]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-03-13T00:00:00"}