{"id":"https://openalex.org/W7151925546","doi":"https://doi.org/10.48550/arxiv.2604.05695","title":"Let Geometry GUIDE: Layer-wise Unrolling of Geometric Priors in Multimodal LLMs","display_name":"Let Geometry GUIDE: Layer-wise Unrolling of Geometric Priors in Multimodal LLMs","publication_year":2026,"publication_date":"2026-04-07","ids":{"openalex":"https://openalex.org/W7151925546","doi":"https://doi.org/10.48550/arxiv.2604.05695"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.05695","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.05695","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":null,"license_id":null,"version":null,"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":"https://doi.org/10.48550/arxiv.2604.05695","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133180882","display_name":"Chongyu Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Wang, Chongyu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133158625","display_name":"Ting Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Ting","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102201210","display_name":"Chunyu Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Chunyu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133191197","display_name":"Xinyu Ning","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ning, Xinyu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133166346","display_name":"Di Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Di","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133166231","display_name":"Hao Tang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tang, Hao","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":6,"corresponding_author_ids":["https://openalex.org/A5133180882"],"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":0.7620000243186951,"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":0.7620000243186951,"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/T10719","display_name":"3D Shape Modeling and Analysis","score":0.056699998676776886,"subfield":{"id":"https://openalex.org/subfields/2206","display_name":"Computational Mechanics"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.02370000071823597,"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/prior-probability","display_name":"Prior probability","score":0.852400004863739},{"id":"https://openalex.org/keywords/geometric-modeling","display_name":"Geometric modeling","score":0.5863000154495239},{"id":"https://openalex.org/keywords/geometric-shape","display_name":"Geometric shape","score":0.5073999762535095},{"id":"https://openalex.org/keywords/geometric-distribution","display_name":"Geometric distribution","score":0.4207000136375427},{"id":"https://openalex.org/keywords/transformation-geometry","display_name":"Transformation geometry","score":0.4083000123500824},{"id":"https://openalex.org/keywords/geometric-transformation","display_name":"Geometric transformation","score":0.4065000116825104},{"id":"https://openalex.org/keywords/geometric-primitive","display_name":"Geometric primitive","score":0.3984000086784363},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.3686000108718872}],"concepts":[{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.852400004863739},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6140000224113464},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6003999710083008},{"id":"https://openalex.org/C104065381","wikidata":"https://www.wikidata.org/wiki/Q1002535","display_name":"Geometric modeling","level":2,"score":0.5863000154495239},{"id":"https://openalex.org/C7305733","wikidata":"https://www.wikidata.org/wiki/Q207961","display_name":"Geometric shape","level":2,"score":0.5073999762535095},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.4244000017642975},{"id":"https://openalex.org/C39947850","wikidata":"https://www.wikidata.org/wiki/Q729523","display_name":"Geometric distribution","level":3,"score":0.4207000136375427},{"id":"https://openalex.org/C32990609","wikidata":"https://www.wikidata.org/wiki/Q306542","display_name":"Transformation geometry","level":2,"score":0.4083000123500824},{"id":"https://openalex.org/C56435381","wikidata":"https://www.wikidata.org/wiki/Q1196371","display_name":"Geometric transformation","level":3,"score":0.4065000116825104},{"id":"https://openalex.org/C181095308","wikidata":"https://www.wikidata.org/wiki/Q1541599","display_name":"Geometric primitive","level":2,"score":0.3984000086784363},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.3686000108718872},{"id":"https://openalex.org/C136520226","wikidata":"https://www.wikidata.org/wiki/Q302814","display_name":"Geometric data analysis","level":2,"score":0.352400004863739},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.3319000005722046},{"id":"https://openalex.org/C108882727","wikidata":"https://www.wikidata.org/wiki/Q2991685","display_name":"Solid modeling","level":2,"score":0.31380000710487366},{"id":"https://openalex.org/C40319758","wikidata":"https://www.wikidata.org/wiki/Q5535477","display_name":"Geometric design","level":2,"score":0.3093999922275543},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.271699994802475},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.26980000734329224},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.26269999146461487},{"id":"https://openalex.org/C26760741","wikidata":"https://www.wikidata.org/wiki/Q160402","display_name":"Perception","level":2,"score":0.2599000036716461},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.25429999828338623},{"id":"https://openalex.org/C141353440","wikidata":"https://www.wikidata.org/wiki/Q182221","display_name":"Fuse (electrical)","level":2,"score":0.25380000472068787}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.05695","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.05695","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.05695","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.05695","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"article"},"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":{"Multimodal":[0],"Large":[1],"Language":[2],"Models":[3],"(MLLMs)":[4],"have":[5,35],"achieved":[6],"remarkable":[7],"progress":[8],"in":[9,76],"2D":[10],"visual":[11,23],"tasks":[12],"but":[13],"still":[14],"exhibit":[15],"limited":[16],"physical":[17],"spatial":[18,168,180,200],"awareness":[19],"when":[20],"processing":[21],"real-world":[22],"streams.":[24],"Recently,":[25],"feed-forward":[26],"geometric":[27,33,70,94,104,125,141,186,212],"foundation":[28],"models,":[29],"which":[30],"implicitly":[31],"extract":[32],"priors,":[34],"provided":[36],"a":[37,92,158,206],"new":[38],"pathway":[39],"to":[40,65,114,148,165],"address":[41],"this":[42,82,143],"issue.":[43],"However,":[44],"existing":[45,195],"geometry-aware":[46],"MLLMs":[47],"are":[48],"predominantly":[49],"constrained":[50],"by":[51],"the":[52,66,77,103,129,133,137,146,151,163,176],"paradigm":[53,208],"of":[54,68,132,139,179],"single":[55],"deep-layer":[56],"extraction":[57],"and":[58,72,121,182,202],"input-level":[59],"fusion.":[60],"This":[61],"flattened":[62],"fusion":[63],"leads":[64],"loss":[67],"local":[69,112],"details":[71],"causes":[73],"semantic":[74],"mismatches":[75],"early":[78,130],"layers.":[79],"To":[80],"break":[81],"bottleneck,":[83],"we":[84,118,156],"propose":[85],"GUIDE":[86,98,192],"(Geometric":[87],"Unrolling":[88],"Inside":[89],"MLLM":[90],"Early-layers),":[91],"progressive":[93],"priors":[95,126,181,213],"injection":[96,138],"framework.":[97],"performs":[99],"multi-level":[100,124],"sampling":[101],"within":[102],"encoder,":[105],"comprehensively":[106],"capturing":[107],"multi-granularity":[108,140],"features":[109],"ranging":[110],"from":[111],"edges":[113],"global":[115],"topologies.":[116],"Subsequently,":[117],"rigorously":[119],"align":[120],"fuse":[122],"these":[123],"step-by-step":[127],"with":[128],"layers":[131],"MLLM.":[134],"Building":[135],"upon":[136],"information,":[142],"design":[144],"guides":[145],"model":[147,164],"progressively":[149],"learn":[150],"2D-to-3D":[152],"transitional":[153],"process.":[154],"Furthermore,":[155],"introduce":[157],"context-aware":[159],"gating":[160],"that":[161,191],"enables":[162],"fetch":[166],"requisite":[167],"cues":[169],"based":[170],"on":[171,197],"current":[172],"semantics,":[173],"thereby":[174],"maximizing":[175],"utilization":[177],"efficiency":[178],"effectively":[183],"suppressing":[184],"redundant":[185],"noise.":[187],"Extensive":[188],"experiments":[189],"demonstrate":[190],"significantly":[193],"outperforms":[194],"baselines":[196],"multiple":[198],"complex":[199],"reasoning":[201],"perception":[203],"tasks,":[204],"establishing":[205],"novel":[207],"for":[209],"integrating":[210],"3D":[211],"into":[214],"large":[215],"models.":[216]},"counts_by_year":[],"updated_date":"2026-04-09T06:13:59.934233","created_date":"2026-04-09T00:00:00"}