{"id":"https://openalex.org/W7208838804","doi":null,"title":"TAP-Path: Task-Adaptive Structural and Token Pruning for Efficient and Trustworthy Pathology Foundation Models","display_name":"TAP-Path: Task-Adaptive Structural and Token Pruning for Efficient and Trustworthy Pathology Foundation Models","publication_year":2026,"publication_date":"2026-09-03","ids":{"openalex":"https://openalex.org/W7208838804"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.04071","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.04071","pdf_url":"https://arxiv.org/pdf/2609.04071","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","indexed_in":["arxiv"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.04071","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5144324280","display_name":"Mehedi Hasan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I5518804","display_name":"BRAC University","ror":"https://ror.org/00sge8677","country_code":"BD","type":"education","lineage":["https://openalex.org/I5518804"]}],"countries":["BD"],"is_corresponding":false,"raw_author_name":"Hasan, Mehedi","raw_affiliation_strings":["Department of Computer Science and Engineering , Brac University , Dhaka , Bangladesh"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science and Engineering , Brac University , Dhaka , Bangladesh","institution_ids":["https://openalex.org/I5518804"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5025044511","display_name":"Ashfak Yeafi","orcid":"https://orcid.org/0009-0007-5084-9191","observed_orcids":["https://orcid.org/0009-0007-5084-9191"]},"institutions":[{"id":"https://openalex.org/I1307585291","display_name":"Khulna University of Engineering and Technology","ror":"https://ror.org/04y58d606","country_code":"BD","type":"education","lineage":["https://openalex.org/I1307585291"]}],"countries":["BD"],"is_corresponding":false,"raw_author_name":"Yeafi, Ashfak","raw_affiliation_strings":["EEE , Khulna University of Engineering & Technology , Khulna , Bangladesh"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"EEE , Khulna University of Engineering & Technology , Khulna , Bangladesh","institution_ids":["https://openalex.org/I1307585291"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100671174","display_name":"Md Khairul Islam","orcid":"https://orcid.org/0000-0002-9125-9573","observed_orcids":["https://orcid.org/0000-0002-9125-9573"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Islam, Md Khairul","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10862","display_name":"AI in cancer detection","score":0.9800000190734863,"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/T10862","display_name":"AI in cancer detection","score":0.9800000190734863,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.014600000344216824,"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"}},{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.002199999988079071,"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/pathology-foundation-models","display_name":"pathology foundation models","score":0.906000018119812},{"id":"https://openalex.org/keywords/histopathology","display_name":"histopathology","score":0.9139999747276306},{"id":"https://openalex.org/keywords/model-compression","display_name":"model compression","score":0.8240000009536743},{"id":"https://openalex.org/keywords/fault-detection","display_name":"fault detection","score":0.5170000195503235},{"id":"https://openalex.org/keywords/cptac","display_name":"CPTAC","score":0.6050000190734863},{"id":"https://openalex.org/keywords/transfer-learning","display_name":"transfer learning","score":0.3869999945163727}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7613999843597412},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.6972000002861023},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5817999839782715},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5241000056266785},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.515999972820282},{"id":"https://openalex.org/C2779696439","wikidata":"https://www.wikidata.org/wiki/Q7512811","display_name":"Signature (topology)","level":2,"score":0.412200003862381},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.4097999930381775},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.39070001244544983},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3806000053882599},{"id":"https://openalex.org/C2777522853","wikidata":"https://www.wikidata.org/wiki/Q5276128","display_name":"Digital pathology","level":2,"score":0.3684999942779541},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.36660000681877136},{"id":"https://openalex.org/C43214815","wikidata":"https://www.wikidata.org/wiki/Q7310987","display_name":"Reliability (semiconductor)","level":3,"score":0.3659999966621399},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.3553999960422516},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.3183000087738037},{"id":"https://openalex.org/C59404180","wikidata":"https://www.wikidata.org/wiki/Q17013334","display_name":"Feature learning","level":2,"score":0.3131999969482422},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.30550000071525574},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2996000051498413},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2827000021934509},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.266400009393692},{"id":"https://openalex.org/C153701036","wikidata":"https://www.wikidata.org/wiki/Q659974","display_name":"Trustworthiness","level":2,"score":0.2624000012874603},{"id":"https://openalex.org/C2779231336","wikidata":"https://www.wikidata.org/wiki/Q7534724","display_name":"Sketch","level":2,"score":0.258899986743927},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2538999915122986}],"mesh":[],"locations_count":1,"locations":[{"id":"pmh:oai:arXiv.org:2609.04071","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.04071","pdf_url":"https://arxiv.org/pdf/2609.04071","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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.04071","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.04071","pdf_url":"https://arxiv.org/pdf/2609.04071","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/W7208838804.pdf","grobid_xml":"https://content.openalex.org/works/W7208838804.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4416952611","https://openalex.org/W7203754287","https://openalex.org/W7129099548","https://openalex.org/W4413969988","https://openalex.org/W4414364929","https://openalex.org/W7213284585","https://openalex.org/W7167677466","https://openalex.org/W4414364083","https://openalex.org/W7171508739","https://openalex.org/W4406443873"],"abstract_inverted_index":{"Pathology":[0],"foundation":[1,205],"models":[2,206],"improve":[3,198],"transferable":[4],"representation":[5],"learning":[6],"for":[7,134,139],"histopathology,":[8],"but":[9],"recent":[10],"gains":[11],"often":[12],"rely":[13],"on":[14,126,173],"encoders":[15],"with":[16,132],"hundreds":[17],"of":[18,20,55,75,81,146,153,202],"millions":[19],"parameters":[21,88],"and":[22,64,79,94,121,137,150,182,194,212],"high":[23],"inference":[24],"cost.":[25],"We":[26],"propose":[27],"TAP-Path,":[28],"a":[29,36,45,65,127,143,166],"task-adaptive":[30,192],"compression":[31],"framework":[32],"that":[33,191],"directly":[34],"restructures":[35],"pretrained":[37],"Virchow2":[38,136],"encoder":[39,87,96],"rather":[40],"than":[41],"distilling":[42],"it":[43],"into":[44],"separate":[46],"student.":[47],"TAP-Path":[48,109,141],"combines":[49],"validation-driven":[50],"transformer-block":[51],"selection,":[52],"physical":[53],"removal":[54],"redundant":[56],"blocks,":[57],"input-adaptive":[58],"patch-token":[59],"pruning,":[60,85],"multi-depth":[61],"feature":[62],"recovery,":[63],"lightweight":[66],"gated":[67],"task":[68],"head.":[69],"The":[70],"final":[71],"model":[72],"retains":[73],"24":[74],"32":[76],"transformer":[77],"blocks":[78],"70%":[80],"patch":[82],"tokens":[83],"after":[84],"reducing":[86],"by":[89,98],"24.96%":[90],"(631.24M":[91],"to":[92,101],"473.70M)":[93],"analytical":[95],"compute":[97],"35.20%":[99],"(340.13G":[100],"220.40G":[102],"FLOPs).":[103],"Across":[104],"three":[105],"task-head":[106],"optimization":[107],"seeds,":[108],"achieved":[110,142],"$87.98":[111],"\\pm":[112,117,123,148,155,179,184],"0.067%$":[113],"test":[114],"accuracy,":[115,120],"$81.26":[116],"0.49%$":[118],"balanced":[119,163,186],"$82.38":[122],"0.48%$":[124],"macro-F1":[125],"32-class":[128],"histopathology":[129],"benchmark,":[130],"compared":[131],"86.89%":[133],"full":[135],"87.67%":[138],"UNI2-h.":[140],"Brier":[144],"score":[145],"$0.1800":[147],"0.0005$":[149],"failure-detection":[151],"AUROC":[152],"$0.9047":[154],"0.0060$.":[156],"A":[157],"validation-only":[158],"rare-aware":[159],"objective":[160],"improved":[161],"rare-class":[162],"accuracy":[164,181],"in":[165],"secondary":[167],"operating":[168],"analysis.":[169],"Frozen":[170],"external":[171,213],"evaluation":[172],"433":[174],"CPTAC":[175],"samples":[176],"yielded":[177],"$91.22":[178],"0.83%$":[180],"$91.10":[183],"0.81%$":[185],"accuracy.":[187],"These":[188],"results":[189],"show":[190],"structural":[193],"token":[195],"sparsification":[196],"can":[197],"the":[199],"accuracy-efficiency":[200],"trade-off":[201],"large":[203],"pathology":[204],"while":[207],"preserving":[208],"reliability":[209],"under":[210],"internal":[211],"evaluation.":[214]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-09-06T00:00:00"}