{"id":"https://openalex.org/W7161727739","doi":"https://doi.org/10.48550/arxiv.2605.18549","title":"Monitoring the Internal Monologue: Probe Trajectories Reveal Reasoning Dynamics","display_name":"Monitoring the Internal Monologue: Probe Trajectories Reveal Reasoning Dynamics","publication_year":2026,"publication_date":"2026-05-18","ids":{"openalex":"https://openalex.org/W7161727739","doi":"https://doi.org/10.48550/arxiv.2605.18549"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.18549","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.18549","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":"article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.18549","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136494612","display_name":"Maciej Chrab\u0105szcz","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chrab\u0105szcz, Maciej","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128113022","display_name":"Aleksander Szymczyk","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Szymczyk, Aleksander","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023066991","display_name":"Marcin Sendera","orcid":"https://orcid.org/0000-0002-8741-6919"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sendera, Marcin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136475663","display_name":"Tomasz Trzci\u0144ski","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Trzci\u0144ski, Tomasz","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5069232913","display_name":"Sebastian Cygert","orcid":"https://orcid.org/0000-0002-4763-8381"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cygert, Sebastian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":5,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.16850000619888306,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.16850000619888306,"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/T11303","display_name":"Bayesian Modeling and Causal Inference","score":0.14229999482631683,"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/T11099","display_name":"Autonomous Vehicle Technology and Safety","score":0.09650000184774399,"subfield":{"id":"https://openalex.org/subfields/2203","display_name":"Automotive Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.6826000213623047},{"id":"https://openalex.org/keywords/trajectory","display_name":"Trajectory","score":0.6304000020027161},{"id":"https://openalex.org/keywords/pooling","display_name":"Pooling","score":0.6176000237464905},{"id":"https://openalex.org/keywords/construct","display_name":"Construct (python library)","score":0.5529999732971191},{"id":"https://openalex.org/keywords/reliability","display_name":"Reliability (semiconductor)","score":0.4973999857902527},{"id":"https://openalex.org/keywords/outcome","display_name":"Outcome (game theory)","score":0.4507000148296356},{"id":"https://openalex.org/keywords/dynamics","display_name":"Dynamics (music)","score":0.42149999737739563}],"concepts":[{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6826000213623047},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6669999957084656},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.6304000020027161},{"id":"https://openalex.org/C70437156","wikidata":"https://www.wikidata.org/wiki/Q7228652","display_name":"Pooling","level":2,"score":0.6176000237464905},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.5529999732971191},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.4973999857902527},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.48399999737739563},{"id":"https://openalex.org/C148220186","wikidata":"https://www.wikidata.org/wiki/Q7111912","display_name":"Outcome (game theory)","level":2,"score":0.4507000148296356},{"id":"https://openalex.org/C145912823","wikidata":"https://www.wikidata.org/wiki/Q113558","display_name":"Dynamics (music)","level":2,"score":0.42149999737739563},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.4041999876499176},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.35690000653266907},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3474999964237213},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.33640000224113464},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.3327000141143799},{"id":"https://openalex.org/C2777488183","wikidata":"https://www.wikidata.org/wiki/Q6900510","display_name":"Safety monitoring","level":2,"score":0.32339999079704285},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.2962999939918518},{"id":"https://openalex.org/C83725634","wikidata":"https://www.wikidata.org/wiki/Q7268699","display_name":"Qualitative reasoning","level":2,"score":0.2818000018596649}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.18549","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.18549","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":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.18549","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.18549","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":"article"},"sustainable_development_goals":[{"score":0.4704759120941162,"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"Reasoning":[1],"Models":[2],"(LRMs)":[3],"introduce":[4],"new":[5],"opportunities":[6],"for":[7,147,217],"safety":[8,193],"monitoring":[9,33,218],"through":[10],"their":[11],"Chain":[12],"of":[13,43,74,123,157,192],"Thought":[14],"(CoT)":[15],"reasoning.":[16],"However,":[17],"CoT":[18,56],"is":[19,88,160],"not":[20],"always":[21],"faithful":[22],"to":[23,45,167,174],"the":[24,40,71,79,94,121,145,155,190],"model's":[25],"final":[26],"output,":[27],"undermining":[28],"its":[29],"reliability":[30],"as":[31,213],"a":[32,60,68,75,99,148,214],"tool.":[34],"To":[35,103],"address":[36],"this,":[37],"we":[38,66,108,196],"investigate":[39],"hidden":[41],"representations":[42],"LRMs":[44],"determine":[46],"whether":[47],"future":[48,85,124],"behavior":[49,87],"can":[50],"be":[51],"predicted":[52],"from":[53,98],"prompt":[54],"and":[55,116,152,163,177,185,194],"representations.":[57],"By":[58],"evaluating":[59],"probe":[61,69,180,211],"at":[62],"each":[63],"generated":[64,141],"token,":[65],"construct":[67],"trajectory,":[70],"continuous":[72],"evolution":[73],"concept's":[76],"probability":[77],"across":[78,189],"reasoning":[80,187],"process.":[81],"We":[82,127],"find":[83],"that":[84,112,198,204],"model":[86,125,142],"more":[89],"distinguishable":[90],"when":[91],"examined":[92],"over":[93],"full":[95],"trajectory":[96,199],"than":[97],"single":[100],"static":[101],"prediction.":[102],"characterize":[104],"these":[105],"temporal":[106],"dynamics,":[107],"extract":[109],"signal-processing":[110],"features":[111,200],"capture":[113],"volatility,":[114],"trend,":[115],"steady-state":[117],"behavior,":[118],"significantly":[119],"improving":[120],"separation":[122],"states.":[126],"also":[128],"present":[129],"two":[130],"methodological":[131],"insights.":[132],"First,":[133],"template-based":[134],"training":[135],"data":[136],"achieves":[137,172],"near-parity":[138],"with":[139],"dynamically":[140],"responses,":[143],"eliminating":[144],"need":[146],"costly":[149],"initial":[150],"inference":[151],"labeling.":[153],"Second,":[154],"choice":[156],"pooling":[158],"operation":[159],"critical:":[161],"average-pooling":[162],"last-token":[164],"methods":[165],"collapse":[166],"near-random":[168],"performance,":[169],"while":[170],"max-pooling":[171],"up":[173],"95%":[175],"AUROC":[176],"yields":[178],"stable":[179],"trajectories.":[181],"Using":[182],"four":[183,186],"datasets":[184],"models":[188],"domains":[191],"mathematics,":[195],"demonstrate":[197],"encode":[201],"task-specific":[202],"dynamics":[203],"improve":[205],"outcome":[206],"separability.":[207],"These":[208],"findings":[209],"establish":[210],"trajectories":[212],"complementary":[215],"framework":[216],"LRM":[219],"behavior.":[220],"Warning:":[221],"This":[222],"article":[223],"contains":[224],"potentially":[225],"harmful":[226],"content.":[227]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-05-20T00:00:00"}