{"id":"https://openalex.org/W7154605890","doi":"https://doi.org/10.48550/arxiv.2604.13061","title":"Token Statistics Reveal Conversational Drift in Multi-turn LLM Interaction","display_name":"Token Statistics Reveal Conversational Drift in Multi-turn LLM Interaction","publication_year":2026,"publication_date":"2026-03-18","ids":{"openalex":"https://openalex.org/W7154605890","doi":"https://doi.org/10.48550/arxiv.2604.13061"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.13061","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13061","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.2604.13061","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133766451","display_name":"Wael Hafez","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hafez, Wael","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5130452956","display_name":"Amir Nazeri","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nazeri, Amir","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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.181099995970726,"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":0.181099995970726,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.17219999432563782,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.11659999936819077,"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/security-token","display_name":"Security token","score":0.652899980545044},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5497000217437744},{"id":"https://openalex.org/keywords/measure","display_name":"Measure (data warehouse)","score":0.5059999823570251},{"id":"https://openalex.org/keywords/consistency","display_name":"Consistency (knowledge bases)","score":0.46799999475479126},{"id":"https://openalex.org/keywords/reliability","display_name":"Reliability (semiconductor)","score":0.4674000144004822},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.44679999351501465},{"id":"https://openalex.org/keywords/semantics","display_name":"Semantics (computer science)","score":0.4009999930858612},{"id":"https://openalex.org/keywords/data-integrity","display_name":"Data integrity","score":0.3968000113964081}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7116000056266785},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.652899980545044},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5497000217437744},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.5059999823570251},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.46799999475479126},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.4674000144004822},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.44679999351501465},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.4009999930858612},{"id":"https://openalex.org/C33762810","wikidata":"https://www.wikidata.org/wiki/Q461671","display_name":"Data integrity","level":2,"score":0.3968000113964081},{"id":"https://openalex.org/C205606062","wikidata":"https://www.wikidata.org/wiki/Q5249645","display_name":"Decoupling (probability)","level":2,"score":0.376800000667572},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.3686999976634979},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3555999994277954},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.3449999988079071},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.3357999920845032},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.30720001459121704},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.28870001435279846},{"id":"https://openalex.org/C140779682","wikidata":"https://www.wikidata.org/wiki/Q210868","display_name":"Sampling (signal processing)","level":3,"score":0.28360000252723694},{"id":"https://openalex.org/C130318100","wikidata":"https://www.wikidata.org/wiki/Q2268914","display_name":"Semantic similarity","level":2,"score":0.2824000120162964},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.28130000829696655},{"id":"https://openalex.org/C200601418","wikidata":"https://www.wikidata.org/wiki/Q2193887","display_name":"Reliability engineering","level":1,"score":0.28130000829696655},{"id":"https://openalex.org/C50942859","wikidata":"https://www.wikidata.org/wiki/Q4967193","display_name":"Rectification","level":3,"score":0.2777999937534332},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.27549999952316284},{"id":"https://openalex.org/C106436119","wikidata":"https://www.wikidata.org/wiki/Q836575","display_name":"Quality assurance","level":3,"score":0.25290000438690186}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.13061","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13061","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.2604.13061","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13061","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":[{"score":0.4175730347633362,"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"language":[1],"models,":[2,131],"LLMs,":[3],"are":[4],"increasingly":[5],"deployed":[6],"in":[7,37,107,142,151,175,199],"multiturn":[8],"settings":[9],"where":[10],"earlier":[11],"responses":[12],"shape":[13],"later":[14],"ones,":[15],"making":[16],"reliability":[17,174],"dependent":[18],"on":[19],"whether":[20],"a":[21,108,134],"conversation":[22],"remains":[23],"consistent":[24],"over":[25],"time.":[26],"When":[27],"this":[28,82],"consistency":[29,62,141],"degrades":[30],"undetected,":[31],"downstream":[32],"decisions":[33],"lose":[34],"their":[35],"grounding":[36],"the":[38,52,92,100,112,156,191],"exchange":[39],"that":[40,59,173,187],"produced":[41],"them.":[42,55],"Yet":[43],"current":[44],"evaluation":[45,198],"methods":[46],"assess":[47],"isolated":[48],"outputs":[49],"rather":[50],"than":[51],"interaction":[53,178],"producing":[54],"Here":[56],"we":[57],"show":[58,172],"conversational":[60,119],"structural":[61,140,188],"can":[63,195],"be":[64,180],"monitored":[65],"directly":[66],"from":[67,190],"token":[68,193],"frequency":[69],"statistics,":[70],"without":[71],"embeddings,":[72],"auxiliary":[73,110],"evaluators":[74],"or":[75],"access":[76],"to":[77,99,182],"model":[78,126],"internals.":[79],"We":[80],"formalize":[81],"signal":[83],"as":[84],"Bipredictability,":[85],"P,":[86],"which":[87],"measures":[88],"shared":[89],"predictability":[90],"across":[91],"context,":[93],"response,":[94],"next":[95],"prompt":[96],"loop":[97],"relative":[98],"turn":[101],"total":[102],"uncertainty,":[103],"and":[104,127,155,164,186],"implement":[105],"it":[106],"lightweight":[109],"architecture,":[111],"Information":[113],"Digital":[114],"Twin,":[115],"IDT.":[116],"Across":[117],"4,574":[118],"turns":[120],"spanning":[121],"34":[122],"conditions,":[123],"one":[124],"student":[125],"three":[128],"frontier":[129],"teacher":[130],"P":[132],"established":[133],"stable":[135],"runtime":[136],"baseline,":[137],"aligned":[138],"with":[139,148,166],"85":[143],"percent":[144,168],"of":[145],"conditions":[146],"but":[147],"semantic":[149,197],"quality":[150,184],"only":[152],"44":[153],"percent,":[154],"IDT":[157],"detected":[158],"all":[159],"tested":[160],"contradictions,":[161],"topic":[162],"shifts":[163],"non-sequiturs":[165],"100":[167],"sensitivity.":[169],"These":[170],"results":[171],"extended":[176],"LLM":[177],"cannot":[179],"reduced":[181],"response":[183],"alone,":[185],"monitoring":[189],"observable":[192],"stream":[194],"complement":[196],"deployment.":[200]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-17T00:00:00"}