{"id":"https://openalex.org/W4416037010","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.339","title":"Multimodal Language Models See Better When They Look Shallower","display_name":"Multimodal Language Models See Better When They Look Shallower","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416037010","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.339"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.emnlp-main.339","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.339","pdf_url":"https://aclanthology.org/2025.emnlp-main.339.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.emnlp-main.339.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5108047473","display_name":"Haoran Chen","orcid":"https://orcid.org/0000-0001-8348-7501"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Haoran Chen","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5049747474","display_name":"Junyan Lin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Junyan Lin","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5006817088","display_name":"Xinghao Chen","orcid":"https://orcid.org/0000-0002-2102-8235"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xinghao Chen","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103194601","display_name":"Yue Fan","orcid":"https://orcid.org/0000-0001-6670-2011"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yue Fan","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5008417257","display_name":"Jianfeng Dong","orcid":"https://orcid.org/0000-0001-5244-3274"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jianfeng Dong","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5070648432","display_name":"Xin Jin","orcid":"https://orcid.org/0000-0001-7746-5955"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xin Jin","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100567767","display_name":"Hui Su","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hui Su","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5103097172","display_name":"Jinlan Fu","orcid":"https://orcid.org/0000-0002-0370-1238"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jinlan Fu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5103000631","display_name":"Xiaoyu Shen","orcid":"https://orcid.org/0000-0002-0217-2469"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiaoyu Shen","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":9,"corresponding_author_ids":["https://openalex.org/A5108047473"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.17321326,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"6688","last_page":"6706"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12031","display_name":"Speech and dialogue systems","score":0.33340001106262207,"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/T12031","display_name":"Speech and dialogue systems","score":0.33340001106262207,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.1281999945640564,"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/T10028","display_name":"Topic Modeling","score":0.058400001376867294,"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/natural-language","display_name":"Natural language","score":0.31360000371932983},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.3068000078201294},{"id":"https://openalex.org/keywords/perspective","display_name":"Perspective (graphical)","score":0.27059999108314514},{"id":"https://openalex.org/keywords/semantics","display_name":"Semantics (computer science)","score":0.2702000141143799},{"id":"https://openalex.org/keywords/field","display_name":"Field (mathematics)","score":0.2549999952316284}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6399999856948853},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4259999990463257},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.4050999879837036},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.31360000371932983},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.3068000078201294},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.2946000099182129},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.27059999108314514},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.2702000141143799},{"id":"https://openalex.org/C9652623","wikidata":"https://www.wikidata.org/wiki/Q190109","display_name":"Field (mathematics)","level":2,"score":0.2549999952316284},{"id":"https://openalex.org/C116834253","wikidata":"https://www.wikidata.org/wiki/Q2039217","display_name":"Identification (biology)","level":2,"score":0.24150000512599945}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.emnlp-main.339","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.339","pdf_url":"https://aclanthology.org/2025.emnlp-main.339.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.emnlp-main.339","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.339","pdf_url":"https://aclanthology.org/2025.emnlp-main.339.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G1121271761","display_name":null,"funder_award_id":"Program","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G2087396116","display_name":null,"funder_award_id":"China","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3085993365","display_name":null,"funder_award_id":"(Grant No.","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3317480652","display_name":null,"funder_award_id":"Science","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G391238517","display_name":null,"funder_award_id":", and","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5994120800","display_name":null,"funder_award_id":"Natural","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G7726157001","display_name":null,"funder_award_id":"Grant No.","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416037010.pdf","grobid_xml":"https://content.openalex.org/works/W4416037010.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Multimodal":[0],"large":[1],"language":[2],"models":[3],"(MLLMs)":[4],"typically":[5],"extract":[6],"visual":[7,49,76,128,168],"features":[8],"from":[9],"the":[10,60,71,163],"final":[11],"layers":[12,39,45,53,86,112,122],"of":[13,43,62,75,97,167],"a":[14,141],"pretrained":[15],"Vision":[16],"Transformer":[17],"(ViT).This":[18],"widespread":[19],"deep-layer":[20],"bias,":[21],"however,":[22],"is":[23],"largely":[24],"driven":[25],"by":[26],"empirical":[27],"convention":[28],"rather":[29],"than":[30],"principled":[31,165],"analysis.While":[32],"prior":[33],"studies":[34],"suggest":[35],"that":[36,109,146,174],"different":[37,41],"ViT":[38,85],"capture":[40],"types":[42],"information-shallower":[44],"focusing":[46],"on":[47,65,126,136],"fine":[48],"details":[50],"and":[51,91,120,133,157],"deeper":[52],"aligning":[54],"more":[55],"closely":[56],"with":[57],"textual":[58],"semantics,":[59],"impact":[61],"this":[63],"variation":[64],"MLLM":[66],"performance":[67],"remains":[68],"underexplored.We":[69],"present":[70],"first":[72,164],"comprehensive":[73],"study":[74,166],"layer":[77,93,169],"selection":[78,170],"for":[79],"MLLMs,":[80,172],"analyzing":[81],"representation":[82],"similarity":[83],"across":[84,101],"to":[87],"establish":[88],"shallow,":[89],"middle,":[90],"deep":[92,111],"groupings.Through":[94],"extensive":[95],"evaluation":[96],"MLLMs":[98,175],"(1.4B-7B":[99],"parameters)":[100],"10":[102],"benchmarks":[103],"encompassing":[104],"60+":[105],"tasks,":[106],"we":[107,139],"find":[108],"while":[110],"excel":[113],"in":[114,171],"semantic-rich":[115],"tasks":[116,129],"like":[117],"OCR,":[118],"shallow":[119],"middle":[121],"significantly":[123],"outperform":[124],"them":[125],"fine-grained":[127],"including":[130],"counting,":[131],"positioning,":[132],"object":[134],"localization.Building":[135],"these":[137],"insights,":[138],"propose":[140],"lightweight":[142],"feature":[143],"fusion":[144,159],"method":[145],"strategically":[147],"incorporates":[148],"shallower":[149],"layers,":[150],"achieving":[151],"consistent":[152],"improvements":[153],"over":[154],"both":[155],"single-layer":[156],"specialized":[158],"baselines.Our":[160],"work":[161],"offers":[162],"showing":[173],"can":[176],"often":[177],"see":[178],"better":[179],"when":[180],"they":[181],"look":[182],"shallower.":[183]},"counts_by_year":[],"updated_date":"2026-04-19T08:26:33.389920","created_date":"2025-11-08T00:00:00"}