{"id":"https://openalex.org/W7204454368","doi":"https://doi.org/10.48550/arxiv.2608.24958","title":"Can We Read the Mind of an Audio LLM? A Verbalizable, Multilingual Middle-Layer Workspace","display_name":"Can We Read the Mind of an Audio LLM? A Verbalizable, Multilingual Middle-Layer Workspace","publication_year":2026,"publication_date":"2026-08-25","ids":{"openalex":"https://openalex.org/W7204454368","doi":"https://doi.org/10.48550/arxiv.2608.24958"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.24958","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24958","pdf_url":"https://arxiv.org/pdf/2608.24958","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.24958","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5148416220","display_name":"Jiajun Fan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Fan, Jiajun","raw_affiliation_strings":["Amazon AGI Foundations","University of Illinois Urbana-Champaign"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon AGI Foundations","institution_ids":[]},{"raw_affiliation_string":"University of Illinois Urbana-Champaign","institution_ids":["https://openalex.org/I157725225"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5148432201","display_name":"Jingyuan Li","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Jingyuan","raw_affiliation_strings":["Amazon AGI Foundations"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon AGI Foundations","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5088032905","display_name":"Prashanth Gurunath Shivakumar","orcid":"https://orcid.org/0000-0003-1632-3309","observed_orcids":["https://orcid.org/0000-0003-1632-3309"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shivakumar, Prashanth Gurunath","raw_affiliation_strings":["Amazon AGI Foundations"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon AGI Foundations","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5148424304","display_name":"Qi Luo","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Luo, Qi","raw_affiliation_strings":["Amazon AGI Foundations"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon AGI Foundations","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5091066062","display_name":"Jia-Hong Huang","orcid":"https://orcid.org/0000-0001-7943-2591","observed_orcids":["https://orcid.org/0000-0001-7943-2591"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Jia-Hong","raw_affiliation_strings":["Amazon AGI Foundations"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon AGI Foundations","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5148467814","display_name":"M. Maruf","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Maruf, M.","raw_affiliation_strings":["Amazon AGI Foundations"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon AGI Foundations","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5076628167","display_name":"Roger Ren","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ren, Roger","raw_affiliation_strings":["Amazon AGI Foundations"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon AGI Foundations","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5022646825","display_name":"Yile Gu","orcid":"https://orcid.org/0000-0003-3662-5056","observed_orcids":["https://orcid.org/0000-0003-3662-5056"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gu, Yile","raw_affiliation_strings":["Amazon AGI Foundations"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon AGI Foundations","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5020350278","display_name":"Rahul Pandey","orcid":"https://orcid.org/0000-0003-2748-0502","observed_orcids":["https://orcid.org/0000-0003-2748-0502"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pandey, Rahul","raw_affiliation_strings":["Amazon AGI Foundations"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon AGI Foundations","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100731237","display_name":"Ge Liu","orcid":"https://orcid.org/0000-0001-9383-5186","observed_orcids":["https://orcid.org/0000-0001-9383-5186"]},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Liu, Ge","raw_affiliation_strings":["University of Illinois Urbana-Champaign"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of Illinois Urbana-Champaign","institution_ids":["https://openalex.org/I157725225"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5109934165","display_name":"Ivan Bulyko","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bulyko, Ivan","raw_affiliation_strings":["Amazon AGI Foundations"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Amazon AGI Foundations","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":3,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.6011000275611877,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.6011000275611877,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.1460999995470047,"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":0.10050000250339508,"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/audio-language-models","display_name":"audio-language models","score":0.9649999737739563},{"id":"https://openalex.org/keywords/logit-lens","display_name":"logit lens","score":0.503000020980835},{"id":"https://openalex.org/keywords/paralinguistic-information","display_name":"paralinguistic information","score":0.6949999928474426},{"id":"https://openalex.org/keywords/activation-patching","display_name":"activation patching","score":0.5690000057220459},{"id":"https://openalex.org/keywords/chain-of-thought-reasoning","display_name":"chain-of-thought reasoning","score":0.6159999966621399}],"concepts":[{"id":"https://openalex.org/C61423126","wikidata":"https://www.wikidata.org/wiki/Q187432","display_name":"Scripting language","level":2,"score":0.7531999945640564},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6951000094413757},{"id":"https://openalex.org/C179926584","wikidata":"https://www.wikidata.org/wiki/Q207714","display_name":"Transcription (linguistics)","level":2,"score":0.6028000116348267},{"id":"https://openalex.org/C554936623","wikidata":"https://www.wikidata.org/wiki/Q199657","display_name":"Reading (process)","level":2,"score":0.5613999962806702},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.3569999933242798},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.3440999984741211},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3061999976634979},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2822999954223633},{"id":"https://openalex.org/C203718221","wikidata":"https://www.wikidata.org/wiki/Q491713","display_name":"Sound (geography)","level":2,"score":0.27790001034736633},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.27480000257492065},{"id":"https://openalex.org/C2776035688","wikidata":"https://www.wikidata.org/wiki/Q1606558","display_name":"Affect (linguistics)","level":2,"score":0.27320000529289246}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.24958","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24958","pdf_url":"https://arxiv.org/pdf/2608.24958","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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.24958","doi":"https://doi.org/10.48550/arxiv.2608.24958","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.24958","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.24958","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.24958","pdf_url":"https://arxiv.org/pdf/2608.24958","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":null,"license_id":null,"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/W7204454368.pdf","grobid_xml":"https://content.openalex.org/works/W7204454368.grobid-xml"},"referenced_works_count":16,"referenced_works":["https://openalex.org/W2135255848","https://openalex.org/W2612431505","https://openalex.org/W4327526719","https://openalex.org/W4385327568","https://openalex.org/W4402671241","https://openalex.org/W4409051416","https://openalex.org/W4413925895","https://openalex.org/W4414946584","https://openalex.org/W4415255153","https://openalex.org/W4417322621","https://openalex.org/W6948017749","https://openalex.org/W7148532570","https://openalex.org/W7161129374","https://openalex.org/W7162776923","https://openalex.org/W7162780508","https://openalex.org/W7169817809"],"related_works":["https://openalex.org/W7221112180","https://openalex.org/W7211909867","https://openalex.org/W7154249048","https://openalex.org/W7134830323","https://openalex.org/W7131863247","https://openalex.org/W7164614510","https://openalex.org/W7213891715","https://openalex.org/W7155062258","https://openalex.org/W7140115484","https://openalex.org/W7204574544"],"abstract_inverted_index":{"An":[0],"audio":[1,159,168,300],"language":[2],"model":[3,33,301],"is":[4,100,129,152,191,226,249,264,293],"a":[5,9,39,43,56,95,118,199,270,286,294],"black":[6],"box":[7],"in":[8,62,65,83,135,214,248],"specific":[10],"way:":[11],"we":[12,50],"see":[13],"what":[14,18,298],"it":[15,19,71,103,225,305],"says,":[16],"never":[17],"works":[20,302],"out":[21,303],"on":[22,94,147,197],"the":[23,32,47,53,66,85,87,90,110,155,162,167,171,179,194,202,205,211,215,232,236,243,246,252,259,267,277,282,289,307],"way":[24,203],"there,":[25],"and":[26,106,113,140,160,182,221,229,255],"chain-of-thought":[27],"monitoring":[28],"helps":[29],"only":[30,276],"if":[31],"writes":[34],"its":[35],"reasoning":[36],"down.":[37],"Reading":[38],"base":[40],"Qwen3-Omni":[41],"with":[42,123],"logit":[44],"lens":[45],"at":[46,138,193],"audio-token":[48],"positions,":[49],"find":[51],"that":[52,178],"answer":[54,256],"to":[55,115,251,258],"spoken":[57],"question":[58],"becomes":[59],"legible":[60],"-":[61,64,117,273,280],"words":[63],"model's":[67,91,163],"middle":[68,216],"layers,":[69],"before":[70,231,304],"emits":[72],"any":[73],"token.":[74],"Five":[75],"findings":[76],"follow.":[77],"(1)":[78],"The":[79,127,188],"readout":[80],"carries":[81],"concepts":[82],"neither":[84],"question,":[86],"options,":[88],"nor":[89],"own":[92,164],"transcription:":[93],"clip":[96,157],"whose":[97],"verbatim":[98],"transcription":[99],"empty":[101],"garbling,":[102],"reconstructs":[104],"Watergate":[105],"scandal,":[107],"passes":[108],"through":[109],"role":[111],"president,":[112],"resolves":[114],"Nixon":[116],"hidden":[119],"multi-hop":[120],"chain,":[121],"read":[122],"no":[124],"chain-of-thought.":[125],"(2)":[126],"content":[128],"language-agnostic:":[130],"one":[131],"audio-inferred":[132],"concept":[133],"surfaces":[134],"several":[136],"scripts":[137],"once,":[139],"38%":[141],"of":[142,201,219,235,297],"top-1":[143],"readouts":[144],"are":[145,309],"Chinese":[146],"English":[148],"inputs.":[149],"(3)":[150],"It":[151],"paralinguistic:":[153],"given":[154],"same":[156],"as":[158,161],"emotion-free":[165],"caption,":[166],"mind":[169],"forms":[170],"sound":[172,247,278],"source,":[173],"speaker":[174],"role,":[175],"or":[176],"affect":[177],"caption":[180],"discards,":[181],"answers":[183],"correctly":[184],"more":[185],"often.":[186],"(4)":[187],"audio-driven":[189,283],"signal":[190,284],"absent":[192],"input,":[195],"turns":[196],"about":[198],"tenth":[200],"into":[204],"network,":[206],"separates":[207],"most":[208],"cleanly":[209],"from":[210,285],"text":[212],"prior":[213,287],"band":[217],"(35-80%":[218],"depth),":[220],"activation":[222],"patching":[223],"shows":[224],"causally":[227],"used":[228],"committed":[230],"last":[233],"fifth":[234],"layers.":[237],"(5)":[238],"Deleting":[239],"single":[240],"layers":[241,254],"maps":[242],"pipeline:":[244],"reading":[245],"localized":[250],"entry":[253],"delivery":[257],"output":[260],"layer,":[261],"while":[262],"retrieval":[263],"distributed":[265],"across":[266],"interior.":[268],"Throughout,":[269],"waveform-swap":[271],"control":[272],"identical":[274],"text,":[275],"changed":[279],"isolates":[281],"over":[288],"printed":[290],"options.":[291],"This":[292],"qualitative":[295],"account":[296],"an":[299],"speaks:":[306],"quantities":[308],"controls,":[310],"not":[311],"benchmark":[312],"scores.":[313]},"counts_by_year":[{"year":2026,"cited_by_count":3}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-08-28T00:00:00"}
