{"id":"https://openalex.org/W7161607687","doi":"https://doi.org/10.48550/arxiv.2605.15913","title":"Towards Generalization of Block Attention via Automatic Segmentation and Block Distillation","display_name":"Towards Generalization of Block Attention via Automatic Segmentation and Block Distillation","publication_year":2026,"publication_date":"2026-05-15","ids":{"openalex":"https://openalex.org/W7161607687","doi":"https://doi.org/10.48550/arxiv.2605.15913"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.15913","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.15913","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":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.15913","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136355695","display_name":"Shuaiyi Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Shuaiyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101677227","display_name":"Zhisong Zhang","orcid":"https://orcid.org/0000-0002-3680-8701"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Zhisong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136364860","display_name":"Yan Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Yan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136427403","display_name":"Lei Zhu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Lei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5003880966","display_name":"Dongyang Ma","orcid":"https://orcid.org/0000-0002-5545-5876"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Dongyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136388342","display_name":"Chenlong Deng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Deng, Chenlong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136436473","display_name":"Yang Deng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Deng, Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136382373","display_name":"Wai Lam","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lam, Wai","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/T10036","display_name":"Advanced Neural Network Applications","score":0.11860000342130661,"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"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.11860000342130661,"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/T12607","display_name":"Personal Information Management and User Behavior","score":0.09700000286102295,"subfield":{"id":"https://openalex.org/subfields/1802","display_name":"Information Systems and Management"},"field":{"id":"https://openalex.org/fields/18","display_name":"Decision Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T12016","display_name":"Web Data Mining and Analysis","score":0.07519999891519547,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/block","display_name":"Block (permutation group theory)","score":0.7373999953269958},{"id":"https://openalex.org/keywords/segmentation","display_name":"Segmentation","score":0.6086000204086304},{"id":"https://openalex.org/keywords/leverage","display_name":"Leverage (statistics)","score":0.5460000038146973},{"id":"https://openalex.org/keywords/partition","display_name":"Partition (number theory)","score":0.4722000062465668},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.40869998931884766},{"id":"https://openalex.org/keywords/heuristic","display_name":"Heuristic","score":0.3747999966144562},{"id":"https://openalex.org/keywords/weighting","display_name":"Weighting","score":0.36169999837875366},{"id":"https://openalex.org/keywords/reuse","display_name":"Reuse","score":0.3605000078678131},{"id":"https://openalex.org/keywords/dropout","display_name":"Dropout (neural networks)","score":0.3382999897003174}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7688000202178955},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.7373999953269958},{"id":"https://openalex.org/C89600930","wikidata":"https://www.wikidata.org/wiki/Q1423946","display_name":"Segmentation","level":2,"score":0.6086000204086304},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.5460000038146973},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.49729999899864197},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.47850000858306885},{"id":"https://openalex.org/C42812","wikidata":"https://www.wikidata.org/wiki/Q1082910","display_name":"Partition (number theory)","level":2,"score":0.4722000062465668},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.40869998931884766},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.3747999966144562},{"id":"https://openalex.org/C183115368","wikidata":"https://www.wikidata.org/wiki/Q856577","display_name":"Weighting","level":2,"score":0.36169999837875366},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.3605000078678131},{"id":"https://openalex.org/C2776145597","wikidata":"https://www.wikidata.org/wiki/Q25339462","display_name":"Dropout (neural networks)","level":2,"score":0.3382999897003174},{"id":"https://openalex.org/C125308379","wikidata":"https://www.wikidata.org/wiki/Q363057","display_name":"Market segmentation","level":2,"score":0.3330000042915344},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.3285999894142151},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3246999979019165},{"id":"https://openalex.org/C70437156","wikidata":"https://www.wikidata.org/wiki/Q7228652","display_name":"Pooling","level":2,"score":0.3125999867916107},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.3059999942779541},{"id":"https://openalex.org/C55166926","wikidata":"https://www.wikidata.org/wiki/Q2892946","display_name":"Oracle","level":2,"score":0.2946999967098236},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.28360000252723694},{"id":"https://openalex.org/C115051666","wikidata":"https://www.wikidata.org/wiki/Q6522493","display_name":"Ranging","level":2,"score":0.27709999680519104},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.27129998803138733},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.2709999978542328},{"id":"https://openalex.org/C61326573","wikidata":"https://www.wikidata.org/wiki/Q1496376","display_name":"Gaussian process","level":3,"score":0.2694000005722046},{"id":"https://openalex.org/C53533937","wikidata":"https://www.wikidata.org/wiki/Q185020","display_name":"Histogram","level":3,"score":0.26409998536109924},{"id":"https://openalex.org/C73555534","wikidata":"https://www.wikidata.org/wiki/Q622825","display_name":"Cluster analysis","level":2,"score":0.2614000141620636}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.15913","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.15913","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":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.15913","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.15913","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":"Preprint"},"sustainable_development_goals":[{"score":0.7974620461463928,"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Block":[0],"attention,":[1,201],"which":[2,131],"processes":[3],"the":[4,41,52,140],"input":[5,45],"as":[6,27],"separate":[7],"blocks":[8,112],"that":[9,59,124,185],"cannot":[10],"attend":[11],"to":[12,18,96,106,138,152,161,172],"one":[13],"another,":[14],"offers":[15],"significant":[16],"potential":[17],"improve":[19],"KV":[20],"cache":[21],"reuse":[22],"in":[23],"long-context":[24],"scenarios":[25],"such":[26],"Retrieval-Augmented":[28],"Generation":[29],"(RAG).":[30],"However,":[31],"its":[32],"broader":[33],"application":[34],"is":[35,125],"hindered":[36],"by":[37],"two":[38],"key":[39],"challenges:":[40],"difficulty":[42],"of":[43,54],"segmenting":[44],"text":[46,91,109],"into":[47,110],"meaningful,":[48],"self-contained":[49],"blocks,":[50,167],"and":[51,72,88,168,182,190,193,205],"inefficiency":[53],"existing":[55],"block":[56,119,129,149,157,159,194,200,210],"fine-tuning":[57],"methods":[58],"risk":[60],"degrading":[61],"performance.":[62],"To":[63],"address":[64],"these,":[65],"we":[66,101,117],"first":[67],"construct":[68],"SemanticSeg,":[69],"a":[70,103,121,133,203],"large":[71],"diverse":[73],"semantic":[74],"segmentation":[75],"dataset":[76],"containing":[77],"over":[78],"30k":[79],"instances":[80],"across":[81,179],"16":[82],"categories-including":[83],"books,":[84],"code,":[85],"web":[86],"text,":[87],"conversations":[89],"with":[90,113],"lengths":[92],"ranging":[93],"from":[94,165],"2k":[95],"32k.":[97],"Using":[98],"this":[99],"dataset,":[100],"train":[102],"lightweight":[104],"segmenter":[105,187],"automatically":[107],"partition":[108],"human-instinct-aligned":[111],"controllable":[114],"granularity.":[115],"Second,":[116],"propose":[118],"distillation,":[120],"training":[122,163],"framework":[123,144],"more":[126],"efficient":[127],"than":[128],"fine-tuning,":[130],"uses":[132],"frozen":[134],"full-attention":[135],"teacher":[136],"model":[137],"guide":[139],"block-attention":[141],"student.":[142],"This":[143],"integrates":[145],"three":[146],"novel":[147],"components:":[148],"sink":[150],"tokens":[151],"mitigate":[153],"information":[154],"loss":[155,170],"at":[156],"boundaries,":[158],"dropout":[160],"leverage":[162],"signals":[164],"all":[166],"token-level":[169],"weighting":[171],"focus":[173],"learning":[174],"on":[175],"block-attention-sensitive":[176],"tokens.":[177],"Experiments":[178],"multiple":[180],"models":[181],"benchmarks":[183],"demonstrate":[184],"our":[186],"outperforms":[188],"heuristic":[189],"statistical":[191],"baselines,":[192],"distillation":[195],"achieves":[196],"near-full-attention":[197],"performance":[198],"under":[199],"establishing":[202],"practical":[204],"scalable":[206],"pathway":[207],"for":[208],"deploying":[209],"attention.":[211]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-19T00:00:00"}