{"id":"https://openalex.org/W7129258056","doi":"https://doi.org/10.48550/arxiv.2602.13315","title":"IDPruner: Harmonizing Importance and Diversity in Visual Token Pruning for MLLMs","display_name":"IDPruner: Harmonizing Importance and Diversity in Visual Token Pruning for MLLMs","publication_year":2026,"publication_date":"2026-02-10","ids":{"openalex":"https://openalex.org/W7129258056","doi":"https://doi.org/10.48550/arxiv.2602.13315"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2602.13315","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":null,"any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5126204057","display_name":"Yifan Tan","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Tan, Yifan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075553304","display_name":"Yifu Sun","orcid":"https://orcid.org/0000-0003-4924-9387"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Yifu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126209242","display_name":"Shirui Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Shirui","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126240028","display_name":"Hong Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Hong","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5077083856","display_name":"Guopan Yu","orcid":"https://orcid.org/0000-0002-5732-0318"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Guanghua","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5059894156","display_name":"Jianchen Zhu","orcid":"https://orcid.org/0000-0002-5988-3704"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Jianchen","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5126224901","display_name":"Yangdong Deng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Deng, Yangdong","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":7,"corresponding_author_ids":["https://openalex.org/A5126204057"],"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":0.8519999980926514,"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":0.8519999980926514,"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/T10028","display_name":"Topic Modeling","score":0.02800000086426735,"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.01899999938905239,"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/security-token","display_name":"Security token","score":0.8230999708175659},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.567300021648407},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.44749999046325684},{"id":"https://openalex.org/keywords/categorization","display_name":"Categorization","score":0.43479999899864197},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.42739999294281006},{"id":"https://openalex.org/keywords/relevance","display_name":"Relevance (law)","score":0.3903999924659729},{"id":"https://openalex.org/keywords/software-deployment","display_name":"Software deployment","score":0.3714999854564667}],"concepts":[{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.8230999708175659},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8087999820709229},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.567300021648407},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5156999826431274},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.45260000228881836},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.44749999046325684},{"id":"https://openalex.org/C94124525","wikidata":"https://www.wikidata.org/wiki/Q912550","display_name":"Categorization","level":2,"score":0.43479999899864197},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.42739999294281006},{"id":"https://openalex.org/C158154518","wikidata":"https://www.wikidata.org/wiki/Q7310970","display_name":"Relevance (law)","level":2,"score":0.3903999924659729},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.3714999854564667},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.3384999930858612},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.30709999799728394},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3010999858379364},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.2971000075340271},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2849000096321106},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.2728999853134155},{"id":"https://openalex.org/C115067241","wikidata":"https://www.wikidata.org/wiki/Q1639854","display_name":"Token passing","level":3,"score":0.25609999895095825}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2602.13315","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},{"id":"doi:10.48550/arxiv.2602.13315","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.13315","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":"article"}],"best_oa_location":{"id":"pmh:doi:10.48550/arxiv.2602.13315","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Multimodal":[0],"Large":[1],"Language":[2],"Models":[3],"(MLLMs)":[4],"have":[5],"demonstrated":[6],"impressive":[7],"capabilities,":[8],"yet":[9],"they":[10],"encounter":[11],"significant":[12],"computational":[13],"bottlenecks":[14],"due":[15],"to":[16,74,104],"the":[17,30,76,90,98,170],"massive":[18],"volume":[19],"of":[20,54,163,169],"visual":[21,24],"tokens.":[22],"Consequently,":[23],"token":[25,31,47,79],"pruning,":[26],"which":[27,96],"substantially":[28],"reduces":[29],"count,":[32],"has":[33],"emerged":[34],"as":[35],"a":[36,57,71,106],"critical":[37],"technique":[38],"for":[39,60],"accelerating":[40],"MLLM":[41],"inference.":[42],"Existing":[43],"approaches":[44],"focus":[45],"on":[46,158],"importance,":[48],"diversity,":[49],"or":[50],"an":[51,178],"intuitive":[52],"combination":[53],"both,":[55],"without":[56,117],"principled":[58],"framework":[59],"their":[61],"optimal":[62],"integration.":[63],"To":[64],"address":[65],"this":[66,86],"issue,":[67],"we":[68,88],"first":[69],"conduct":[70,133],"systematic":[72],"analysis":[73],"characterize":[75],"trade-off":[77],"between":[78,109],"importance":[80],"and":[81,92,126,140,149,155,172],"semantic":[82],"diversity.":[83],"Guided":[84],"by":[85],"analysis,":[87],"propose":[89],"\\textbf{I}mportance":[91],"\\textbf{D}iversity":[93],"Pruner":[94],"(\\textbf{IDPruner}),":[95],"leverages":[97],"Maximal":[99],"Marginal":[100],"Relevance":[101],"(MMR)":[102],"algorithm":[103],"achieve":[105],"Pareto-optimal":[107],"balance":[108],"these":[110],"two":[111],"objectives.":[112],"Crucially,":[113],"our":[114],"method":[115],"operates":[116],"requiring":[118],"attention":[119],"maps,":[120],"ensuring":[121],"full":[122],"compatibility":[123],"with":[124],"FlashAttention":[125],"efficient":[127],"deployment":[128],"via":[129],"one-shot":[130],"pruning.":[131],"We":[132],"extensive":[134],"experiments":[135],"across":[136,152],"various":[137],"model":[138],"architectures":[139,154],"multimodal":[141],"benchmarks,":[142],"demonstrating":[143],"that":[144],"IDPruner":[145,160],"achieves":[146],"state-of-the-art":[147],"performance":[148,165],"superior":[150],"generalization":[151],"diverse":[153],"tasks.":[156],"Notably,":[157],"Qwen2.5-VL-7B-Instruct,":[159],"retains":[161],"95.18\\%":[162],"baseline":[164],"when":[166],"pruning":[167,181],"75\\%":[168],"tokens,":[171],"still":[173],"maintains":[174],"86.40\\%":[175],"even":[176],"under":[177],"extreme":[179],"90\\%":[180],"ratio.":[182],"Our":[183],"code":[184],"is":[185],"available":[186],"at":[187],"https://github.com/Tencent/AngelSlim.":[188]},"counts_by_year":[],"updated_date":"2026-04-04T16:13:02.066488","created_date":"2026-02-18T00:00:00"}