{"id":"https://openalex.org/W4416036764","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.470","title":"Look Again, Think Slowly: Enhancing Visual Reflection in Vision-Language Models","display_name":"Look Again, Think Slowly: Enhancing Visual Reflection in Vision-Language Models","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416036764","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.470"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.emnlp-main.470","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.470","pdf_url":"https://aclanthology.org/2025.emnlp-main.470.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.470.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100622420","display_name":"Jian Pu","orcid":"https://orcid.org/0000-0002-8487-9467"},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Pu Jian","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101499870","display_name":"Junhong Wu","orcid":"https://orcid.org/0000-0002-2679-1627"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Junhong Wu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100726958","display_name":"Wei Sun","orcid":"https://orcid.org/0000-0002-8801-6060"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei Sun","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100337673","display_name":"Chen Wang","orcid":"https://orcid.org/0000-0003-2922-5345"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen Wang","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5030954686","display_name":"Shuo Ren","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shuo Ren","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5111205847","display_name":"Jiajun Zhang","orcid":"https://orcid.org/0009-0002-9712-9953"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jiajun Zhang","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5100622420"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":1.33,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.8621765,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":97,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"9262","last_page":"9281"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.7890999913215637,"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":0.7890999913215637,"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/T11148","display_name":"Language, Metaphor, and Cognition","score":0.012600000016391277,"subfield":{"id":"https://openalex.org/subfields/3205","display_name":"Experimental and Cognitive Psychology"},"field":{"id":"https://openalex.org/fields/32","display_name":"Psychology"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T12031","display_name":"Speech and dialogue systems","score":0.011900000274181366,"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/reflection","display_name":"Reflection (computer programming)","score":0.5483999848365784},{"id":"https://openalex.org/keywords/perspective","display_name":"Perspective (graphical)","score":0.30799999833106995},{"id":"https://openalex.org/keywords/visualization","display_name":"Visualization","score":0.30550000071525574},{"id":"https://openalex.org/keywords/focus","display_name":"Focus (optics)","score":0.2865000069141388},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.2770000100135803}],"concepts":[{"id":"https://openalex.org/C65682993","wikidata":"https://www.wikidata.org/wiki/Q1056451","display_name":"Reflection (computer programming)","level":2,"score":0.5483999848365784},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4828999936580658},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3659999966621399},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3409000039100647},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.30799999833106995},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.30550000071525574},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.2865000069141388},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.2770000100135803},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.2630000114440918},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.26249998807907104},{"id":"https://openalex.org/C119657128","wikidata":"https://www.wikidata.org/wiki/Q11633","display_name":"Photography","level":2,"score":0.2549000084400177},{"id":"https://openalex.org/C26760741","wikidata":"https://www.wikidata.org/wiki/Q160402","display_name":"Perception","level":2,"score":0.2513999938964844}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.emnlp-main.470","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.470","pdf_url":"https://aclanthology.org/2025.emnlp-main.470.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.470","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.470","pdf_url":"https://aclanthology.org/2025.emnlp-main.470.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":[],"funders":[{"id":"https://openalex.org/F4320321133","display_name":"Chinese Academy of Sciences","ror":"https://ror.org/034t30j35"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416036764.pdf","grobid_xml":"https://content.openalex.org/works/W4416036764.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Recent":[0],"advances":[1],"in":[2,31,165],"text-only":[3],"\"slow-thinking\"":[4],"reasoning":[5,20,41,87,102,113,134,146],"have":[6],"prompted":[7],"efforts":[8],"to":[9,13,38,61,132],"transfer":[10,24],"this":[11,71],"capability":[12],"vision-language":[14],"models":[15,21],"(VLMs),":[16],"for":[17,90,95],"training":[18],"visual":[19,34,45,56,62,83,119,123,137,145,157,160,166],"(VRMs).However,":[22],"such":[23],"faces":[25],"critical":[26],"challenges:":[27],"Effective":[28],"\"slow":[29],"thinking\"":[30],"VRMs":[32,53],"requires":[33],"reflection,":[35,57],"the":[36,40],"ability":[37],"check":[39],"process":[42],"based":[43,85,125,135],"on":[44,86,136,156],"information.Through":[46],"quantitative":[47],"analysis,":[48],"we":[49,73,99],"observe":[50],"that":[51,108],"current":[52],"exhibit":[54],"limited":[55],"as":[58],"their":[59],"attention":[60,124],"information":[63,158],"diminishes":[64],"rapidly":[65],"with":[66],"longer":[67],"generated":[68],"responses.To":[69],"address":[70],"challenge,":[72],"propose":[74],"a":[75,122,150],"new":[76],"VRM":[77],"Reflection-V":[78,139,148],"1":[79],",":[80],"which":[81],"enhances":[82],"reflection":[84,120,167],"data":[88,103],"construction":[89],"cold-start":[91,116],"and":[92,112,152],"reward":[93,126],"design":[94],"reinforcement":[96],"learning":[97,117],"(RL).Firstly,":[98],"construct":[100],"vision-centered":[101],"by":[104],"leveraging":[105],"an":[106],"agent":[107],"interacts":[109],"between":[110],"VLMs":[111],"LLMs,":[114],"enabling":[115],"of":[118],"patterns.Secondly,":[121],"model":[127],"is":[128],"employed":[129],"during":[130,159],"RL":[131],"encourage":[133],"information.Therefore,":[138],"demonstrates":[140],"significant":[141],"improvements":[142],"across":[143],"multiple":[144],"benchmarks.Furthermore,":[147],"maintains":[149],"stronger":[151],"more":[153],"consistent":[154],"reliance":[155],"reasoning,":[161],"indicating":[162],"effective":[163],"enhancement":[164],"capabilities.":[168]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-03-20T23:20:44.827607","created_date":"2025-11-08T00:00:00"}