{"id":"https://openalex.org/W7162570738","doi":"https://doi.org/10.48550/arxiv.2605.27194","title":"Not All Tokens Matter Equally: Dynamic In-context Vector Distillation with Decisive-Token Supervision for Long-form Medical Report Generation","display_name":"Not All Tokens Matter Equally: Dynamic In-context Vector Distillation with Decisive-Token Supervision for Long-form Medical Report Generation","publication_year":2026,"publication_date":"2026-05-26","ids":{"openalex":"https://openalex.org/W7162570738","doi":"https://doi.org/10.48550/arxiv.2605.27194"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.27194","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.27194","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.2605.27194","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137102130","display_name":"Ning Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Ning","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137132732","display_name":"Rui Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Rui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137141916","display_name":"Xinkun Lin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Xinkun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137165581","display_name":"Weixing Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Weixing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137127906","display_name":"Jinxi Xiang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiang, Jinxi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137117227","display_name":"Tao Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Tao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137161357","display_name":"Lina Yao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yao, Lina","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137132086","display_name":"Mingjie Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Mingjie","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/T10028","display_name":"Topic Modeling","score":0.6520000100135803,"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/T10028","display_name":"Topic Modeling","score":0.6520000100135803,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.14560000598430634,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.07890000194311142,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/decoding-methods","display_name":"Decoding methods","score":0.7827000021934509},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.5138000249862671},{"id":"https://openalex.org/keywords/natural-language-generation","display_name":"Natural language generation","score":0.5101000070571899},{"id":"https://openalex.org/keywords/encoding","display_name":"Encoding (memory)","score":0.46070000529289246},{"id":"https://openalex.org/keywords/distillation","display_name":"Distillation","score":0.4343999922275543},{"id":"https://openalex.org/keywords/fidelity","display_name":"Fidelity","score":0.4246000051498413},{"id":"https://openalex.org/keywords/quality","display_name":"Quality (philosophy)","score":0.40880000591278076},{"id":"https://openalex.org/keywords/signal","display_name":"SIGNAL (programming language)","score":0.3953000009059906}],"concepts":[{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.7827000021934509},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7470999956130981},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.5138000249862671},{"id":"https://openalex.org/C2776187449","wikidata":"https://www.wikidata.org/wiki/Q1513879","display_name":"Natural language generation","level":3,"score":0.5101000070571899},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.46070000529289246},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.4343999922275543},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4327999949455261},{"id":"https://openalex.org/C2776459999","wikidata":"https://www.wikidata.org/wiki/Q2119376","display_name":"Fidelity","level":2,"score":0.4246000051498413},{"id":"https://openalex.org/C2779530757","wikidata":"https://www.wikidata.org/wiki/Q1207505","display_name":"Quality (philosophy)","level":2,"score":0.40880000591278076},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.3953000009059906},{"id":"https://openalex.org/C2779662365","wikidata":"https://www.wikidata.org/wiki/Q5416694","display_name":"Event (particle physics)","level":2,"score":0.3758000135421753},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.34860000014305115},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3476000130176544},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.34119999408721924},{"id":"https://openalex.org/C186644900","wikidata":"https://www.wikidata.org/wiki/Q194152","display_name":"Parsing","level":2,"score":0.3276999890804291},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.3098999857902527},{"id":"https://openalex.org/C149364088","wikidata":"https://www.wikidata.org/wiki/Q185917","display_name":"Translation (biology)","level":4,"score":0.30889999866485596},{"id":"https://openalex.org/C113364801","wikidata":"https://www.wikidata.org/wiki/Q26674","display_name":"High fidelity","level":2,"score":0.30869999527931213},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.30550000071525574},{"id":"https://openalex.org/C111219384","wikidata":"https://www.wikidata.org/wiki/Q6954384","display_name":"NIST","level":2,"score":0.295199990272522},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2825999855995178},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.28200000524520874},{"id":"https://openalex.org/C159877910","wikidata":"https://www.wikidata.org/wiki/Q2202883","display_name":"Autoregressive model","level":2,"score":0.2750000059604645},{"id":"https://openalex.org/C107418235","wikidata":"https://www.wikidata.org/wiki/Q1520565","display_name":"Human multitasking","level":2,"score":0.2700999975204468},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.251800000667572},{"id":"https://openalex.org/C2985684807","wikidata":"https://www.wikidata.org/wiki/Q1513879","display_name":"Text generation","level":2,"score":0.2515999972820282},{"id":"https://openalex.org/C154030694","wikidata":"https://www.wikidata.org/wiki/Q1436074","display_name":"Fractionating column","level":3,"score":0.25029999017715454}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.27194","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.27194","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.2605.27194","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.27194","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":[{"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4","score":0.4740985035896301}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Distilling":[0],"demonstration":[1],"effects":[2],"into":[3],"hidden-space":[4],"interventions":[5],"offers":[6],"a":[7,27,37,122],"lightweight":[8],"alternative":[9],"to":[10,33,136,186],"full":[11],"finetuning.":[12],"However,":[13],"existing":[14],"multimodal":[15],"variants":[16],"are":[17,55,71,162],"mostly":[18],"evaluated":[19],"on":[20,192,233],"short-form":[21],"tasks,":[22],"where":[23],"outputs":[24,54],"end":[25],"after":[26],"few":[28],"tokens.":[29],"Extending":[30],"these":[31,137],"methods":[32,210],"long-form":[34,53,128],"generation":[35,77,130],"exposes":[36],"fundamental":[38],"yet":[39],"underexamined":[40],"limitation:":[41],"token-level":[42],"distillation":[43,124],"implicitly":[44],"treats":[45],"all":[46,227],"output":[47,69],"tokens":[48,65,82,86,151],"as":[49,188],"equally":[50],"informative,":[51],"but":[52],"dominated":[56],"by":[57,113,144],"high-frequency":[58],"template":[59],"and":[60,91,106,152,160,194,213,223],"grammatical":[61],"tokens,":[62],"while":[63,230],"the":[64,92,111,146,153,183,208,219],"that":[66,87,96,126,157,203],"actually":[67],"determine":[68,88],"quality":[70],"sparsely":[72],"distributed.":[73],"In":[74],"medical":[75,199],"report":[76,129],"(MRG),":[78],"two":[79,132,198],"such":[80],"decisive":[81],"stand":[83],"out:":[84],"pathology-related":[85,150],"diagnostic":[89],"content,":[90],"end-of-sequence":[93],"(EOS)":[94],"event":[95],"determines":[97],"termination.":[98],"Both":[99],"receive":[100],"insufficient":[101],"supervision":[102,140,142],"under":[103],"uniform":[104],"cross-entropy,":[105],"autoregressive":[107],"decoding":[108,170,189],"further":[109],"compounds":[110],"problem":[112],"drifting":[114],"away":[115],"from":[116],"teacher-forced":[117],"trajectories.":[118],"We":[119],"propose":[120],"DIVE,":[121],"frozen-backbone":[123],"framework":[125],"addresses":[127],"through":[131],"complementary":[133],"mechanisms":[134],"matched":[135],"failures.":[138],"Decisive-token":[139],"restores":[141],"balance":[143],"upweighting":[145],"cross-entropy":[147],"contribution":[148],"of":[149],"EOS":[154],"event,":[155],"ensuring":[156],"content":[158],"fidelity":[159],"termination":[161],"learned":[163],"during":[164],"training":[165],"rather":[166],"than":[167],"imposed":[168],"at":[169],"time.":[171],"State-conditioned":[172],"dynamic":[173],"steering":[174],"replaces":[175],"fixed":[176],"open-loop":[177],"residuals":[178],"with":[179,197],"hidden-state-dependent":[180],"adapters,":[181],"allowing":[182],"injected":[184],"signal":[185],"adapt":[187],"drifts.":[190],"Experiments":[191],"MIMIC-CXR":[193],"CheXpert":[195],"Plus":[196],"VLM":[200],"backbones":[201],"show":[202],"DIVE":[204],"consistently":[205],"ranks":[206],"among":[207],"strongest":[209],"across":[211],"lexical":[212],"clinical-proxy":[214],"metrics.":[215],"Our":[216],"method":[217],"achieves":[218],"best":[220],"BLEU-4,":[221],"ROUGE-L,":[222],"RadGraph":[224],"F1":[225],"in":[226],"dataset--backbone":[228],"settings,":[229],"remaining":[231],"competitive":[232],"coarse":[234],"label-level":[235],"CheXbert":[236],"F1.":[237]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-28T00:00:00"}