{"id":"https://openalex.org/W7165389488","doi":"https://doi.org/10.48550/arxiv.2606.20075","title":"What Makes Effective Supervision in Latent Chain-of-Thought: An Information-Theoretic Analysis","display_name":"What Makes Effective Supervision in Latent Chain-of-Thought: An Information-Theoretic Analysis","publication_year":2026,"publication_date":"2026-06-18","ids":{"openalex":"https://openalex.org/W7165389488","doi":"https://doi.org/10.48550/arxiv.2606.20075"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.20075","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.20075","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":null,"license_id":null,"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.2606.20075","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5138974977","display_name":"Xinghao Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Xinghao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139007925","display_name":"Chak Tou Leong","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Leong, Chak Tou","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138988865","display_name":"Wenjin Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Wenjin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138981052","display_name":"Jian Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Jian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138992032","display_name":"Wenjie Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Wenjie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139018550","display_name":"Xiaoyu Shen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shen, Xiaoyu","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/T11883","display_name":"Embodied and Extended Cognition","score":0.4296000003814697,"subfield":{"id":"https://openalex.org/subfields/2805","display_name":"Cognitive Neuroscience"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}},"topics":[{"id":"https://openalex.org/T11883","display_name":"Embodied and Extended Cognition","score":0.4296000003814697,"subfield":{"id":"https://openalex.org/subfields/2805","display_name":"Cognitive Neuroscience"},"field":{"id":"https://openalex.org/fields/28","display_name":"Neuroscience"},"domain":{"id":"https://openalex.org/domains/1","display_name":"Life Sciences"}},{"id":"https://openalex.org/T11431","display_name":"Action Observation and Synchronization","score":0.09799999743700027,"subfield":{"id":"https://openalex.org/subfields/3207","display_name":"Social Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.05389999970793724,"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/latent-variable","display_name":"Latent variable","score":0.5508000254631042},{"id":"https://openalex.org/keywords/probabilistic-latent-semantic-analysis","display_name":"Probabilistic latent semantic analysis","score":0.4336000084877014},{"id":"https://openalex.org/keywords/fidelity","display_name":"Fidelity","score":0.4250999987125397},{"id":"https://openalex.org/keywords/perspective","display_name":"Perspective (graphical)","score":0.41990000009536743},{"id":"https://openalex.org/keywords/trajectory","display_name":"Trajectory","score":0.3871000111103058},{"id":"https://openalex.org/keywords/semantics","display_name":"Semantics (computer science)","score":0.3716000020503998},{"id":"https://openalex.org/keywords/path","display_name":"Path (computing)","score":0.3612000048160553},{"id":"https://openalex.org/keywords/latent-semantic-analysis","display_name":"Latent semantic analysis","score":0.3546000123023987},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.3440999984741211}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6068000197410583},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5627999901771545},{"id":"https://openalex.org/C51167844","wikidata":"https://www.wikidata.org/wiki/Q4422623","display_name":"Latent variable","level":2,"score":0.5508000254631042},{"id":"https://openalex.org/C112933361","wikidata":"https://www.wikidata.org/wiki/Q2845258","display_name":"Probabilistic latent semantic analysis","level":2,"score":0.4336000084877014},{"id":"https://openalex.org/C2776459999","wikidata":"https://www.wikidata.org/wiki/Q2119376","display_name":"Fidelity","level":2,"score":0.4250999987125397},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.41990000009536743},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.3871000111103058},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3734000027179718},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.3716000020503998},{"id":"https://openalex.org/C2777735758","wikidata":"https://www.wikidata.org/wiki/Q817765","display_name":"Path (computing)","level":2,"score":0.3612000048160553},{"id":"https://openalex.org/C170133592","wikidata":"https://www.wikidata.org/wiki/Q1806883","display_name":"Latent semantic analysis","level":2,"score":0.3546000123023987},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.3440999984741211},{"id":"https://openalex.org/C65965080","wikidata":"https://www.wikidata.org/wiki/Q1806885","display_name":"Latent variable model","level":3,"score":0.3294000029563904},{"id":"https://openalex.org/C115086926","wikidata":"https://www.wikidata.org/wiki/Q17004651","display_name":"Causal reasoning","level":3,"score":0.32829999923706055},{"id":"https://openalex.org/C107848011","wikidata":"https://www.wikidata.org/wiki/Q4680756","display_name":"Adaptive reasoning","level":4,"score":0.3140999972820282},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.3095000088214874},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.30079999566078186},{"id":"https://openalex.org/C20162079","wikidata":"https://www.wikidata.org/wiki/Q1151406","display_name":"Case-based reasoning","level":2,"score":0.29910001158714294},{"id":"https://openalex.org/C89288958","wikidata":"https://www.wikidata.org/wiki/Q7301504","display_name":"Reasoning system","level":2,"score":0.28139999508857727},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.2806999981403351},{"id":"https://openalex.org/C152139883","wikidata":"https://www.wikidata.org/wiki/Q252973","display_name":"Mutual information","level":2,"score":0.27390000224113464},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2605000138282776},{"id":"https://openalex.org/C40506919","wikidata":"https://www.wikidata.org/wiki/Q7452469","display_name":"Sequence learning","level":2,"score":0.25850000977516174},{"id":"https://openalex.org/C148220186","wikidata":"https://www.wikidata.org/wiki/Q7111912","display_name":"Outcome (game theory)","level":2,"score":0.25769999623298645},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.2500999867916107}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.20075","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.20075","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.20075","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.20075","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Latent":[0,44,134],"Chain-of-Thought":[1],"(CoT)":[2],"internalizes":[3],"reasoning":[4,16,21,86,110,147,155,175],"within":[5],"continuous":[6],"hidden":[7],"states,":[8],"offering":[9],"a":[10,55,116,151,170],"promising":[11],"alternative":[12],"to":[13,36],"verbose":[14],"discrete":[15],"traces.":[17],"However,":[18],"robust":[19],"latent":[20,33,69,98,143,165,174],"remains":[22],"difficult":[23],"because":[24],"outcome":[25],"supervision":[26,75,176],"provides":[27,115],"weak":[28],"learning":[29],"signals":[30],"and":[31,50,64,88,145,177],"leaves":[32],"trajectories":[34,144],"prone":[35],"semantic":[37,94,119],"drift.":[38],"In":[39],"this":[40,52],"work,":[41],"we":[42,130],"analyze":[43],"CoT":[45],"from":[46,180],"an":[47],"information-theoretic":[48],"perspective":[49],"identify":[51],"failure":[53],"as":[54],"dual":[56],"collapse:":[57],"gradient":[58],"attenuation":[59],"along":[60],"the":[61,68,93,97,109,132,139,159,164],"optimization":[62],"path":[63],"representational":[65],"drift":[66],"in":[67,163],"space.":[70],"We":[71],"further":[72],"decompose":[73],"process":[74],"into":[76],"two":[77],"complementary":[78],"dimensions:":[79],"Trajectory":[80],"Supervision,":[81,90],"which":[82,91,137],"injects":[83],"dense":[84],"stepwise":[85],"signals,":[87],"Space":[89],"preserves":[92,123],"structure":[95],"of":[96],"manifold.":[99],"Our":[100,187],"analysis":[101],"shows":[102],"that":[103,121],"rigid":[104],"geometric":[105,181],"compression":[106],"can":[107],"collapse":[108],"space,":[111],"whereas":[112],"generative":[113],"reconstruction":[114],"more":[117],"flexible":[118],"anchor":[120],"better":[122],"information":[124,141,160,185],"capacity.":[125],"To":[126],"measure":[127],"these":[128],"effects,":[129],"introduce":[131],"Unified":[133],"Probe":[135],"(ULP),":[136],"quantifies":[138],"mutual":[140,184],"between":[142],"explicit":[146],"steps.":[148],"Experiments":[149],"reveal":[150],"clear":[152],"Information-Performance":[153],"Binding:":[154],"accuracy":[156],"depends":[157],"on":[158],"fidelity":[161],"preserved":[162],"chain.":[166],"These":[167],"findings":[168],"provide":[169],"principled":[171],"framework":[172],"for":[173],"suggest":[178],"shifting":[179],"imitation":[182],"toward":[183],"maximization.":[186],"code":[188],"is":[189],"available":[190],"at":[191],"\\href{https://github.com/EIT-NLP/Supervision-in-Latent-CoT}{this":[192],"repository}.":[193]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-20T00:00:00"}