{"id":"https://openalex.org/W7162770711","doi":"https://doi.org/10.48550/arxiv.2605.30231","title":"Beyond 3D VQAs: Injecting 3D Spatial Priors into Vision-Language Models for Enhanced Geometric Reasoning","display_name":"Beyond 3D VQAs: Injecting 3D Spatial Priors into Vision-Language Models for Enhanced Geometric Reasoning","publication_year":2026,"publication_date":"2026-05-28","ids":{"openalex":"https://openalex.org/W7162770711","doi":"https://doi.org/10.48550/arxiv.2605.30231"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.30231","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.30231","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,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.30231","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5111130446","display_name":"Chun-Hsiao Yeh","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yeh, Chun-Hsiao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5030632256","display_name":"Shengyi Qian","orcid":"https://orcid.org/0000-0003-0262-2412"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qian, Shengyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5059021916","display_name":"Manchen Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Manchen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137335416","display_name":"Yi Ma","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Yi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5002114219","display_name":"Joseph Tighe","orcid":"https://orcid.org/0000-0002-0716-8119"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tighe, Joseph","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5064761222","display_name":"Fanyi Xiao","orcid":"https://orcid.org/0000-0002-9839-1139"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiao, Fanyi","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":0.9904000163078308,"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":0.9904000163078308,"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":0.0020000000949949026,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.0013000000035390258,"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/prior-probability","display_name":"Prior probability","score":0.8055999875068665},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.5809000134468079},{"id":"https://openalex.org/keywords/overfitting","display_name":"Overfitting","score":0.5480999946594238},{"id":"https://openalex.org/keywords/boosting","display_name":"Boosting (machine learning)","score":0.4366999864578247},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.4049000144004822},{"id":"https://openalex.org/keywords/encoder","display_name":"Encoder","score":0.40389999747276306},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.3776000142097473},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.3628999888896942},{"id":"https://openalex.org/keywords/consistency","display_name":"Consistency (knowledge bases)","score":0.3555999994277954}],"concepts":[{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.8055999875068665},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6798999905586243},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6776999831199646},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.5809000134468079},{"id":"https://openalex.org/C22019652","wikidata":"https://www.wikidata.org/wiki/Q331309","display_name":"Overfitting","level":3,"score":0.5480999946594238},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.4366999864578247},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.4049000144004822},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.40389999747276306},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4000000059604645},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.3939000070095062},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3776000142097473},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.3628999888896942},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.3555999994277954},{"id":"https://openalex.org/C32990609","wikidata":"https://www.wikidata.org/wiki/Q306542","display_name":"Transformation geometry","level":2,"score":0.3450999855995178},{"id":"https://openalex.org/C27511587","wikidata":"https://www.wikidata.org/wiki/Q2178623","display_name":"Spatial relation","level":2,"score":0.3443000018596649},{"id":"https://openalex.org/C56435381","wikidata":"https://www.wikidata.org/wiki/Q1196371","display_name":"Geometric transformation","level":3,"score":0.3425000011920929},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.33480000495910645},{"id":"https://openalex.org/C2911011789","wikidata":"https://www.wikidata.org/wiki/Q130741","display_name":"Hallucinating","level":2,"score":0.3337000012397766},{"id":"https://openalex.org/C159620131","wikidata":"https://www.wikidata.org/wiki/Q1938983","display_name":"Spatial analysis","level":2,"score":0.32350000739097595},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.305400013923645},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.2921999990940094},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C88626702","wikidata":"https://www.wikidata.org/wiki/Q1128903","display_name":"Continuation","level":2,"score":0.289000004529953},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.2791999876499176},{"id":"https://openalex.org/C2780297707","wikidata":"https://www.wikidata.org/wiki/Q4895393","display_name":"Landmark","level":2,"score":0.27320000529289246},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.2685999870300293},{"id":"https://openalex.org/C205711294","wikidata":"https://www.wikidata.org/wiki/Q176953","display_name":"Rendering (computer graphics)","level":2,"score":0.26010000705718994},{"id":"https://openalex.org/C155512373","wikidata":"https://www.wikidata.org/wiki/Q287450","display_name":"Residual","level":2,"score":0.2526000142097473}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.30231","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.30231","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,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.30231","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.30231","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,"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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Vision-Language":[0],"Models":[1],"(VLMs)":[2],"often":[3,33],"struggle":[4],"with":[5,16,95,218],"robust":[6],"3D":[7,17,29,122,198,221],"spatial":[8,44,182,222],"reasoning.":[9,223],"Prevailing":[10],"methods":[11],"that":[12,42,67,132,148,204],"rely":[13],"on":[14,109,180,186,191,196],"fine-tuning":[15],"visual":[18,30],"question-answering":[19],"(VQA)":[20],"datasets":[21],"may":[22],"overfit":[23],"dataset-specific":[24],"biases,":[25],"while":[26,116,168],"integrating":[27],"specialized":[28],"encoders":[31],"is":[32,93,139,210],"inflexible":[34],"and":[35,92,162,189,213],"cumbersome.":[36],"In":[37],"this":[38,153],"paper,":[39],"we":[40],"argue":[41],"genuine":[43],"understanding":[45],"should":[46],"emerge":[47],"from":[48,55,102,206],"learning":[49,205],"fundamental":[50,207],"geometric":[51,123,208],"priors,":[52],"not":[53],"only":[54],"high-level":[56],"VQA":[57,199],"supervision.":[58],"We":[59,145],"propose":[60],"GASP":[61,77],"(Geometric-Aware":[62],"Spatial":[63],"Priors),":[64],"a":[65,79,85,96,106,117,129,211],"framework":[66],"injects":[68],"these":[69],"priors":[70,209],"directly":[71],"into":[72],"the":[73],"LLM's":[74],"transformer":[75],"layers.":[76],"employs":[78],"small":[80],"correspondence":[81,136,158],"head,":[82],"applied":[83],"as":[84],"deep":[86],"supervision":[87,120],"signal":[88],"across":[89],"all":[90,193],"layers,":[91],"trained":[94],"dual":[97],"objective":[98],"leveraging":[99],"ground-truth":[100,110],"geometry":[101],"large-scale":[103],"video":[104],"scenes:":[105],"contrastive":[107],"loss":[108],"point":[111],"correspondences":[112],"enforces":[113],"2D":[114],"view-invariance,":[115],"depth":[118],"consistency":[119],"resolves":[121],"ambiguities.":[124],"Our":[125,201],"analysis":[126],"first":[127],"provides":[128],"diagnostic":[130],"showing":[131],"standard":[133],"VLMs'":[134],"internal":[135,174],"matching":[137],"accuracy":[138],"very":[140],"low":[141],"(often":[142],"below":[143,171],"5%).":[144],"then":[146],"demonstrate":[147],"our":[149],"training":[150,195],"substantially":[151],"improves":[152],"behavior,":[154],"boosting":[155],"peak":[156],"layer-wise":[157],"to":[159,177],"over":[160,164],"70%":[161],"maintaining":[163],"85%":[165],"temporal":[166],"robustness":[167],"baselines":[169],"remain":[170],"5%.":[172],"These":[173],"improvements":[175],"translate":[176],"significant":[178],"gains":[179],"downstream":[181],"benchmarks":[183],"including":[184],"+18.2%":[185],"All-Angles":[187],"Bench":[188],"+29.0%":[190],"VSI-Bench,":[192],"without":[194],"any":[197],"data.":[200],"findings":[202],"indicate":[203],"promising":[212],"generalizable":[214],"pathway":[215],"towards":[216],"VLMs":[217],"more":[219],"reliable":[220]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-05-30T00:00:00"}