{"id":"https://openalex.org/W7204181294","doi":"https://doi.org/10.48550/arxiv.2608.22332","title":"Mechanistic Interpretability of Chain-of-Thought Reasoning via Sequential Activation Patching","display_name":"Mechanistic Interpretability of Chain-of-Thought Reasoning via Sequential Activation Patching","publication_year":2026,"publication_date":"2026-08-23","ids":{"openalex":"https://openalex.org/W7204181294","doi":"https://doi.org/10.48550/arxiv.2608.22332"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.22332","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.22332","pdf_url":"https://arxiv.org/pdf/2608.22332","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.22332","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5148049285","display_name":"Murat Dura","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I120217413","display_name":"Izmir Institute of Technology","ror":"https://ror.org/03stptj97","country_code":"TR","type":"education","lineage":["https://openalex.org/I120217413"]}],"countries":["TR"],"is_corresponding":false,"raw_author_name":"Dura, Murat","raw_affiliation_strings":["Department of Computer Engineering \u0130zmir Institute of Technology \u0130zmir , T\u00fcrkiye"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Engineering \u0130zmir Institute of Technology \u0130zmir , T\u00fcrkiye","institution_ids":["https://openalex.org/I120217413"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5148164078","display_name":"Serkan \u00d6zt\u00fcrk","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I120217413","display_name":"Izmir Institute of Technology","ror":"https://ror.org/03stptj97","country_code":"TR","type":"education","lineage":["https://openalex.org/I120217413"]}],"countries":["TR"],"is_corresponding":false,"raw_author_name":"\u00d6zt\u00fcrk, Serkan","raw_affiliation_strings":["Department of Computer Engineering \u0130zmir Institute of Technology \u0130zmir , T\u00fcrkiye"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Engineering \u0130zmir Institute of Technology \u0130zmir , T\u00fcrkiye","institution_ids":["https://openalex.org/I120217413"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5036394895","display_name":"Selma Tek\u0131r","orcid":"https://orcid.org/0000-0002-0488-9682","observed_orcids":["https://orcid.org/0000-0002-0488-9682"]},"institutions":[{"id":"https://openalex.org/I120217413","display_name":"Izmir Institute of Technology","ror":"https://ror.org/03stptj97","country_code":"TR","type":"education","lineage":["https://openalex.org/I120217413"]}],"countries":["TR"],"is_corresponding":false,"raw_author_name":"Tekir, Selma","raw_affiliation_strings":["Department of Computer Engineering \u0130zmir Institute of Technology \u0130zmir , T\u00fcrkiye"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Engineering \u0130zmir Institute of Technology \u0130zmir , T\u00fcrkiye","institution_ids":["https://openalex.org/I120217413"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I120217413"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.8720999956130981,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.8720999956130981,"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":0.10199999809265137,"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":0.010499999858438969,"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/chain-of-thought-prompting","display_name":"chain-of-thought prompting","score":0.9380000233650208},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.843999981880188},{"id":"https://openalex.org/keywords/activation-patching","display_name":"activation patching","score":0.8119999766349792},{"id":"https://openalex.org/keywords/attention-heads","display_name":"attention heads","score":0.6919999718666077},{"id":"https://openalex.org/keywords/mechanistic-interpretability","display_name":"mechanistic interpretability","score":0.8100000023841858},{"id":"https://openalex.org/keywords/multi-head-attention","display_name":"multi-head attention","score":0.6420000195503235},{"id":"https://openalex.org/keywords/reasoning-paths","display_name":"reasoning paths","score":0.5770000219345093},{"id":"https://openalex.org/keywords/part-of-speech-analysis","display_name":"part-of-speech analysis","score":0.6579999923706055}],"concepts":[{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.9194999933242798},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7445999979972839},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5652999877929688},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5354999899864197},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.36629998683929443},{"id":"https://openalex.org/C198082294","wikidata":"https://www.wikidata.org/wiki/Q3399648","display_name":"Position (finance)","level":2,"score":0.3659000098705292},{"id":"https://openalex.org/C89611455","wikidata":"https://www.wikidata.org/wiki/Q6804646","display_name":"Mechanism (biology)","level":2,"score":0.35899999737739563},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.32659998536109924},{"id":"https://openalex.org/C169900460","wikidata":"https://www.wikidata.org/wiki/Q2200417","display_name":"Cognition","level":2,"score":0.31520000100135803},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2824999988079071},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.26829999685287476},{"id":"https://openalex.org/C188147891","wikidata":"https://www.wikidata.org/wiki/Q147638","display_name":"Cognitive science","level":1,"score":0.2529999911785126}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.22332","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.22332","pdf_url":"https://arxiv.org/pdf/2608.22332","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.22332","doi":"https://doi.org/10.48550/arxiv.2608.22332","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.22332","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,"listed_in":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.22332","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.22332","pdf_url":"https://arxiv.org/pdf/2608.22332","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,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204181294.pdf","grobid_xml":"https://content.openalex.org/works/W7204181294.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4392340654","https://openalex.org/W4416036346","https://openalex.org/W4414930565","https://openalex.org/W4399417123","https://openalex.org/W7166875430","https://openalex.org/W7102624217","https://openalex.org/W7124175571","https://openalex.org/W4399837031","https://openalex.org/W4405094836","https://openalex.org/W4417097545"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"demonstrate":[4],"remarkable":[5],"problem-solving":[6],"capabilities":[7],"when":[8],"guided":[9],"by":[10],"Chain-of-Thought":[11],"(CoT)":[12],"prompting,":[13],"yet":[14],"the":[15,35,109,128],"internal":[16],"mechanisms":[17],"underlying":[18],"these":[19,71],"improvements":[20],"remain":[21],"poorly":[22],"understood.":[23],"In":[24],"this":[25,77],"work,":[26],"we":[27,79],"investigate":[28],"where":[29],"CoT-related":[30],"causal":[31],"effects":[32,97],"emerge":[33],"across":[34,91],"generated":[36,56],"reasoning":[37,52],"trajectory":[38],"and":[39,94,119,138,150],"which":[40],"attention":[41],"heads":[42,130],"carry":[43],"signals":[44],"that":[45,86,127],"contribute":[46],"to":[47,69,107],"final-answer":[48],"computation.":[49,165],"Because":[50],"CoT":[51],"unfolds":[53],"over":[54],"multiple":[55],"tokens,":[57],"standard":[58],"activation":[59,83,121],"patching":[60,84],"at":[61],"a":[62,81],"single":[63],"static":[64],"token":[65,92],"position":[66],"is":[67],"insufficient":[68],"characterize":[70],"temporally":[72],"distributed":[73,113,159],"effects.":[74],"To":[75],"address":[76],"limitation,":[78],"introduce":[80,103],"sequential":[82],"framework":[85],"traces":[87],"CoT-conditioned":[88,164],"attention-head":[89],"activations":[90],"positions":[93],"aggregates":[95],"their":[96],"using":[98],"Part-of-Speech-guided":[99],"analysis.":[100],"We":[101],"further":[102],"Sequential":[104],"Multi-Head":[105],"Patching":[106],"evaluate":[108],"joint":[110],"contribution":[111],"of":[112],"head":[114],"sets,":[115],"together":[116],"with":[117,163],"cross-question":[118],"random":[120],"controls.":[122],"Targeted":[123],"zero-ablation":[124],"experiments":[125],"show":[126],"identified":[129],"are":[131],"functionally":[132],"important":[133],"for":[134,158],"successful":[135],"answer":[136,146],"generation":[137],"affect":[139],"several":[140],"overlapping":[141],"mechanisms,":[142],"including":[143],"reasoning-trajectory":[144],"maintenance,":[145],"anchoring,":[147],"exemplar-target":[148],"separation,":[149],"numerical":[151],"generation.":[152],"Overall,":[153],"our":[154],"results":[155],"provide":[156],"evidence":[157],"reasoning-support":[160],"sub-circuits":[161],"associated":[162]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-26T00:00:00"}