{"id":"https://openalex.org/W7164845559","doi":"https://doi.org/10.1145/3805622.3810807","title":"Beyond Inconsistent Reasoning: Intermediate Process Direct Preference Optimization on Small Multi-modal Large Language Models for Visual Question Answering","display_name":"Beyond Inconsistent Reasoning: Intermediate Process Direct Preference Optimization on Small Multi-modal Large Language Models for Visual Question Answering","publication_year":2026,"publication_date":"2026-06-15","ids":{"openalex":"https://openalex.org/W7164845559","doi":"https://doi.org/10.1145/3805622.3810807"},"language":null,"primary_location":{"id":"doi:10.1145/3805622.3810807","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3805622.3810807","pdf_url":null,"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 2026 International Conference on Multimedia Retrieval","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3805622.3810807","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101237510","display_name":"Yongzhu Miao","orcid":null},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yongzhu Miao","raw_affiliation_strings":["School of Computer Science, National University of Defense Technology, Changsha, Hunan, China"],"raw_orcid":"https://orcid.org/0009-0008-8127-499X","affiliations":[{"raw_affiliation_string":"School of Computer Science, National University of Defense Technology, Changsha, Hunan, China","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124150746","display_name":"Puzhen Su","orcid":null},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Puzhen Su","raw_affiliation_strings":["School of Computer Science, National University of Defense Technology, Changsha, Hunan, China"],"raw_orcid":"https://orcid.org/0009-0001-7613-1192","affiliations":[{"raw_affiliation_string":"School of Computer Science, National University of Defense Technology, Changsha, Hunan, China","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126835490","display_name":"Haoran Yin","orcid":null},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Haoran Yin","raw_affiliation_strings":["School of Computer Science, National University of Defense Technology, Changsha, Hunan, China"],"raw_orcid":"https://orcid.org/0009-0003-7471-0996","affiliations":[{"raw_affiliation_string":"School of Computer Science, National University of Defense Technology, Changsha, Hunan, China","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100376019","display_name":"Shasha Li","orcid":"https://orcid.org/0000-0002-6508-5119"},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shasha Li","raw_affiliation_strings":["School of Computer Science, National University of Defense Technology, Changsha, Hunan, China"],"raw_orcid":"https://orcid.org/0000-0002-6508-5119","affiliations":[{"raw_affiliation_string":"School of Computer Science, National University of Defense Technology, Changsha, Hunan, China","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047593277","display_name":"Jintao Tang","orcid":"https://orcid.org/0000-0002-8802-3906"},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jintao Tang","raw_affiliation_strings":["School of Computer Science, National University of Defense Technology, Changsha, Hunan, China"],"raw_orcid":"https://orcid.org/0000-0002-8802-3906","affiliations":[{"raw_affiliation_string":"School of Computer Science, National University of Defense Technology, Changsha, Hunan, China","institution_ids":["https://openalex.org/I170215575"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5020937767","display_name":"Ting Wang","orcid":"https://orcid.org/0000-0002-7780-2330"},"institutions":[{"id":"https://openalex.org/I170215575","display_name":"National University of Defense Technology","ror":"https://ror.org/05d2yfz11","country_code":"CN","type":"education","lineage":["https://openalex.org/I170215575"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ting Wang","raw_affiliation_strings":["School of Computer Science, National University of Defense Technology, Changsha, Hunan, China"],"raw_orcid":"https://orcid.org/0000-0002-7780-2330","affiliations":[{"raw_affiliation_string":"School of Computer Science, National University of Defense Technology, Changsha, Hunan, China","institution_ids":["https://openalex.org/I170215575"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":6,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.93858172,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"787","last_page":"796"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11714","display_name":"Multimodal Machine Learning Applications","score":0.9955999851226807,"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":0.9955999851226807,"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":0.000699999975040555,"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/T11273","display_name":"Advanced Graph Neural Networks","score":0.00039999998989515007,"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/overfitting","display_name":"Overfitting","score":0.7648000121116638},{"id":"https://openalex.org/keywords/consistency","display_name":"Consistency (knowledge bases)","score":0.7121000289916992},{"id":"https://openalex.org/keywords/preference","display_name":"Preference","score":0.6502000093460083},{"id":"https://openalex.org/keywords/categorization","display_name":"Categorization","score":0.527899980545044},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.5041999816894531},{"id":"https://openalex.org/keywords/preference-elicitation","display_name":"Preference elicitation","score":0.44830000400543213}],"concepts":[{"id":"https://openalex.org/C22019652","wikidata":"https://www.wikidata.org/wiki/Q331309","display_name":"Overfitting","level":3,"score":0.7648000121116638},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.7121000289916992},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.675000011920929},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.6502000093460083},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6157000064849854},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5461000204086304},{"id":"https://openalex.org/C94124525","wikidata":"https://www.wikidata.org/wiki/Q912550","display_name":"Categorization","level":2,"score":0.527899980545044},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.5041999816894531},{"id":"https://openalex.org/C2777868144","wikidata":"https://www.wikidata.org/wiki/Q7239817","display_name":"Preference elicitation","level":3,"score":0.44830000400543213},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.4465000033378601},{"id":"https://openalex.org/C181204326","wikidata":"https://www.wikidata.org/wiki/Q7239820","display_name":"Preference learning","level":3,"score":0.3734999895095825},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.36739999055862427},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.3644999861717224},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.2939000129699707},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.2644999921321869},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.25130000710487366}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3805622.3810807","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3805622.3810807","pdf_url":null,"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 2026 International Conference on Multimedia Retrieval","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3805622.3810807","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3805622.3810807","pdf_url":null,"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 2026 International Conference on Multimedia Retrieval","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":16,"referenced_works":["https://openalex.org/W1933349210","https://openalex.org/W2560730294","https://openalex.org/W2963890019","https://openalex.org/W4402671816","https://openalex.org/W4402716477","https://openalex.org/W4402727405","https://openalex.org/W4404782590","https://openalex.org/W4409269797","https://openalex.org/W4412886751","https://openalex.org/W4412888149","https://openalex.org/W4412944664","https://openalex.org/W4413147124","https://openalex.org/W4415798312","https://openalex.org/W4415799075","https://openalex.org/W7160031695","https://openalex.org/W7160207210"],"related_works":[],"abstract_inverted_index":{"Visual":[0],"Question":[1],"Answering":[2],"(VQA)":[3],"has":[4],"increasingly":[5],"shifted":[6],"from":[7,32],"direct":[8],"prediction":[9],"to":[10,125,171,227],"step-by-step":[11],"reasoning,":[12],"driven":[13],"by":[14,92,123,149],"advances":[15],"in":[16,202,210],"Multi-modal":[17,23],"Large":[18],"Language":[19],"Models":[20],"(MLLMs)":[21],"and":[22,61,73,84,86,89,112,119,179,187],"Chain-of-Thought":[24],"(MCoT).":[25],"However,":[26],"small":[27],"MLLMs":[28],"(S-MLLMs)":[29],"often":[30],"suffer":[31],"a":[33,55,98],"\"Faithfulness":[34],"Gap\",":[35],"where":[36],"they":[37],"arrive":[38],"at":[39],"correct":[40],"answers":[41],"with":[42,102,163],"inconsistent":[43],"or":[44],"hallucinatory":[45],"intermediate":[46,78,114,152,203],"reasoning":[47,221],"processes.":[48],"In":[49],"this":[50,134],"work,":[51],"we":[52,95,136],"first":[53],"conduct":[54],"systematic":[56],"diagnostic":[57],"of":[58,129],"MCoT":[59,169],"consistency":[60,204],"identify":[62],"two":[63,191],"prevalent":[64],"patterns":[65],"that":[66,110,182,200,215],"undermine":[67],"conventional":[68],"trajectory-level":[69],"optimization:":[70],"Intermediate":[71,138],"Error":[72],"Inconsistent":[74],"Reasoning.":[75],"We":[76],"categorize":[77],"errors":[79,115],"into":[80],"Caption":[81],"Errors\u2014Hallucination,":[82],"Contradiction,":[83],"Irrelevance,":[85],"Reasoning":[87],"Errors\u2014Divergence":[88],"Spuriousness.":[90],"Driven":[91],"these":[93],"findings,":[94],"curate":[96],"MCoT-SP,":[97],"Stage-wise":[99],"Preference":[100,141],"dataset":[101],"severity-aware":[103],"annotations":[104],"constructed":[105],"through":[106],"an":[107],"automated":[108],"pipeline":[109],"detects":[111],"corrects":[113],"using":[116],"entailment-based":[117],"checks":[118],"LLM-based":[120],"correction,":[121],"followed":[122],"post-verification":[124],"ensure":[126],"the":[127,130,207,219],"reliability":[128],"preference":[131,147,156,161],"pairs.":[132],"Leveraging":[133],"resource,":[135],"propose":[137],"Process":[139],"Direct":[140],"Optimization":[142],"(IP-DPO),":[143],"which":[144],"performs":[145],"stage-wise":[146],"optimization":[148],"treating":[150],"each":[151],"stage":[153],"as":[154],"independent":[155],"units,":[157],"while":[158],"jointly":[159],"integrating":[160],"learning":[162],"supervised":[164],"fine-tuning":[165],"on":[166,177],"preferred":[167],"complete":[168],"responses":[170],"preserve":[172],"global":[173],"coherence.":[174],"Extensive":[175],"evaluations":[176],"LLaVA-Next":[178],"Qwen2.5-VL":[180],"demonstrate":[181],"IP-DPO":[183,216],"outperforms":[184],"standard":[185],"DPO":[186],"SFT":[188],"baselines":[189],"across":[190],"benchmarks":[192],"spanning":[193],"six":[194],"VQA":[195],"datasets.":[196],"Our":[197],"analysis":[198],"indicates":[199],"improvements":[201,209],"substantially":[205],"exceed":[206],"corresponding":[208],"final":[211],"answer":[212],"accuracy,":[213],"revealing":[214],"primarily":[217],"improves":[218],"underlying":[220],"process":[222],"rather":[223],"than":[224],"merely":[225],"overfitting":[226],"target":[228],"labels.":[229]},"counts_by_year":[],"updated_date":"2026-06-16T07:37:23.134862","created_date":"2026-06-16T00:00:00"}