{"id":"https://openalex.org/W7204185199","doi":"https://doi.org/10.48550/arxiv.2608.22963","title":"Buried in Textual Debt: Context Pruning with Visual Evidence Preservation for MLLM Agents","display_name":"Buried in Textual Debt: Context Pruning with Visual Evidence Preservation for MLLM Agents","publication_year":2026,"publication_date":"2026-08-24","ids":{"openalex":"https://openalex.org/W7204185199","doi":"https://doi.org/10.48550/arxiv.2608.22963"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.22963","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.22963","pdf_url":"https://arxiv.org/pdf/2608.22963","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.22963","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133525704","display_name":"Yuchen Huang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Yuchen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5148075234","display_name":"Sijia Li","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Sijia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5148107174","display_name":"Jun Zhang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Jun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5093945868","display_name":"Yi R. Fung","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fung, Yi R.","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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9980000257492065,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9980000257492065,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.0003000000142492354,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.0003000000142492354,"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/multimodal-large-language-models","display_name":"multimodal large language models","score":0.9150000214576721},{"id":"https://openalex.org/keywords/kullback-leibler-divergence","display_name":"Kullback-Leibler divergence","score":0.8019999861717224},{"id":"https://openalex.org/keywords/reverse-kl-divergence","display_name":"reverse KL divergence","score":0.7369999885559082},{"id":"https://openalex.org/keywords/self-distillation","display_name":"self-distillation","score":0.8870000243186951}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7537999749183655},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6873999834060669},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.5993000268936157},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.5789999961853027},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.5325000286102295},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.4952000081539154},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.47920000553131104},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.47870001196861267},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.460099995136261},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.39410001039505005},{"id":"https://openalex.org/C194648553","wikidata":"https://www.wikidata.org/wiki/Q1364774","display_name":"Spare part","level":2,"score":0.3824999928474426},{"id":"https://openalex.org/C173801870","wikidata":"https://www.wikidata.org/wiki/Q201413","display_name":"Heuristic","level":2,"score":0.3596999943256378},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.3215999901294708},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.3018999993801117},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.30070000886917114},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2980000078678131},{"id":"https://openalex.org/C56461940","wikidata":"https://www.wikidata.org/wiki/Q970687","display_name":"Eye tracking","level":2,"score":0.29159998893737793},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.2913999855518341},{"id":"https://openalex.org/C189430467","wikidata":"https://www.wikidata.org/wiki/Q7293293","display_name":"Ranking (information retrieval)","level":2,"score":0.2831000089645386},{"id":"https://openalex.org/C183322885","wikidata":"https://www.wikidata.org/wiki/Q17007702","display_name":"Context model","level":3,"score":0.2827000021934509},{"id":"https://openalex.org/C2776505523","wikidata":"https://www.wikidata.org/wiki/Q4785468","display_name":"Plan (archaeology)","level":2,"score":0.26649999618530273},{"id":"https://openalex.org/C2779525943","wikidata":"https://www.wikidata.org/wiki/Q1187300","display_name":"Grammaticality","level":3,"score":0.2597000002861023},{"id":"https://openalex.org/C2780103172","wikidata":"https://www.wikidata.org/wiki/Q1309721","display_name":"Visual Objects","level":3,"score":0.25920000672340393},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.2572000026702881},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.25529998540878296},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.25369998812675476}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.22963","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.22963","pdf_url":"https://arxiv.org/pdf/2608.22963","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.22963","doi":"https://doi.org/10.48550/arxiv.2608.22963","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.22963","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":"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.22963","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.22963","pdf_url":"https://arxiv.org/pdf/2608.22963","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","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":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7172451058","https://openalex.org/W7202110368","https://openalex.org/W7150939104","https://openalex.org/W7166822115","https://openalex.org/W4416036089","https://openalex.org/W7154468764","https://openalex.org/W7163077331","https://openalex.org/W7166090344","https://openalex.org/W7162430321","https://openalex.org/W7206149524"],"abstract_inverted_index":{"Multimodal":[0],"Large":[1],"Language":[2],"Models":[3],"(MLLMs)":[4],"are":[5],"increasingly":[6],"deployed":[7],"as":[8,95],"multi-step":[9,153],"agents,":[10],"where":[11],"explicit":[12],"reasoning":[13,44,84,170],"supports":[14],"task":[15],"decomposition":[16],"and":[17,34,111,148,186],"tool":[18],"coordination":[19],"but":[20],"also":[21],"accumulates":[22],"self-generated":[23,190],"text.":[24],"Over":[25],"long":[26],"trajectories,":[27],"this":[28],"text":[29,69],"can":[30,56],"dominate":[31],"the":[32,105,109,123,127,136,159],"context":[33],"suppress":[35],"visual":[36,49,72,154,184],"evidence,":[37],"creating":[38],"textual":[39,179],"debt.":[40],"We":[41,74,133],"observe":[42],"that":[43,177],"becomes":[45],"redundant":[46,68],"once":[47],"task-relevant":[48],"evidence":[50,185],"is":[51],"grounded,":[52],"while":[53,166],"stale":[54],"hypotheses":[55],"misguide":[57],"later":[58],"inference":[59],"when":[60],"grounding":[61],"remains":[62],"uncertain.":[63],"Pruning":[64],"must":[65],"therefore":[66],"remove":[67],"without":[70,129],"discarding":[71],"evidence.":[73],"propose":[75],"SPARE,":[76],"a":[77,91],"Kullback-Leibler":[78],"(KL)-guided":[79],"framework":[80],"for":[81],"pruning":[82,164],"accumulated":[83],"in":[85],"multimodal":[86],"tool-use":[87,155],"agents.":[88],"SPARE":[89,157],"uses":[90],"compact":[92,144],"task-state":[93],"summary":[94,124],"privileged":[96],"diagnostic":[97],"context.":[98],"For":[99],"each":[100],"candidate":[101],"segment,":[102],"it":[103],"replays":[104],"same":[106],"model":[107],"under":[108],"original":[110],"summary-conditioned":[112],"contexts.":[113],"Reverse-KL":[114],"divergence":[115],"from":[116],"on-policy":[117],"self-distillation":[118],"(OPSD)":[119],"then":[120],"tests":[121],"whether":[122],"sufficiently":[125],"covers":[126],"segment":[128],"disrupting":[130],"future":[131],"reasoning.":[132],"further":[134],"fine-tune":[135],"summarizer":[137],"with":[138],"supervised":[139],"fine-tuning":[140],"(SFT),":[141],"enabling":[142],"more":[143,149],"summaries,":[145],"broader":[146],"coverage,":[147],"aggressive":[150],"pruning.":[151],"Across":[152],"benchmarks,":[156],"achieves":[158],"highest":[160],"average":[161],"accuracy":[162],"among":[163],"methods":[165],"removing":[167],"37.89-64.58\\%":[168],"of":[169],"tokens.":[171],"This":[172],"favorable":[173],"accuracy-context":[174],"trade-off":[175],"shows":[176],"reducing":[178],"dominance":[180],"restores":[181],"reliance":[182],"on":[183,189],"mitigates":[187],"over-conditioning":[188],"language.":[191]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-26T00:00:00"}