{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:3BCWRAFNCLVNTQV55HKVPBOZPC","short_pith_number":"pith:3BCWRAFN","schema_version":"1.0","canonical_sha256":"d8456880ad12ead9c2bde9d55785d97885049292fb006e9fe678c8865d21f5db","source":{"kind":"arxiv","id":"2605.16911","version":1},"attestation_state":"computed","paper":{"title":"VGGT-Occ: Geometry-Grounded and Density-Aware Gated Fusion for 3D Occupancy Prediction","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Embedding camera geometry into every attention and fusion step produces more accurate 3D semantic occupancy from multi-view images.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Danwei Wang, Fangjinhua Wang, Hesheng Wang, Hongming Shen, Junyi Ma, Rui Wang, Tianchen Deng, Xun Chen","submitted_at":"2026-05-16T09:51:04Z","abstract_excerpt":"3D semantic occupancy prediction requires accurate 2D-to-3D feature lifting, yet current methods restrict camera geometry to initial projections. Subsequent operations like offset learning, attention weighting, and cross-camera aggregation remain geometry-agnostic, ignoring essential physical constraints. We propose VGGT-Occ, a framework that embeds geometric tokens throughout the entire pipeline. We introduce Projection-Aware Deformable Attention (PA-DA) to inject geometry into all attention stages. PA-DA projects 3D offsets back to image planes and leverages the projection Jacobian as an add"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2605.16911","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-16T09:51:04Z","cross_cats_sorted":[],"title_canon_sha256":"77eeea72c02c88cbd3d75aff2d1ddc552ed1cf35d00707d1dd97b03ccaff796a","abstract_canon_sha256":"1353cf93b475b7f2ca6d0e31faac0017f55055833e5f8becd4e7bc431ccea2fa"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:29.790787Z","signature_b64":"3FqEKD27pgEmM2RmUccQntam83yMGCq627D9TGr2XmUVemnvr8pTwJrn8jHEHQ7vnQbqdX/XJuFjWwii/e3FBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d8456880ad12ead9c2bde9d55785d97885049292fb006e9fe678c8865d21f5db","last_reissued_at":"2026-05-20T00:03:29.789963Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:29.789963Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"VGGT-Occ: Geometry-Grounded and Density-Aware Gated Fusion for 3D Occupancy Prediction","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Embedding camera geometry into every attention and fusion step produces more accurate 3D semantic occupancy from multi-view images.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Danwei Wang, Fangjinhua Wang, Hesheng Wang, Hongming Shen, Junyi Ma, Rui Wang, Tianchen Deng, Xun Chen","submitted_at":"2026-05-16T09:51:04Z","abstract_excerpt":"3D semantic occupancy prediction requires accurate 2D-to-3D feature lifting, yet current methods restrict camera geometry to initial projections. Subsequent operations like offset learning, attention weighting, and cross-camera aggregation remain geometry-agnostic, ignoring essential physical constraints. We propose VGGT-Occ, a framework that embeds geometric tokens throughout the entire pipeline. We introduce Projection-Aware Deformable Attention (PA-DA) to inject geometry into all attention stages. PA-DA projects 3D offsets back to image planes and leverages the projection Jacobian as an add"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"By embedding geometric tokens throughout the pipeline with Projection-Aware Deformable Attention that projects 3D offsets and uses the projection Jacobian as an additive bias, plus a view-quality semantic gate and sequential coarse-to-fine gated fusion, VGGT-Occ achieves 33.00% IoU and 21.08% mIoU on SurroundOcc-nuScenes, outperforming prior methods with only ~41M trainable parameters in the occupancy head.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That projecting 3D offsets back to image planes and adding the projection Jacobian as a bias term will reliably suppress unreliable observations, and that the view-quality semantic gate will correctly enforce cross-view consistency without introducing new errors in feature integration.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"VGGT-Occ embeds geometric tokens via PA-DA and uses sequential coarse-to-fine gated fusion to reach 33.00% IoU and 21.08% mIoU on SurroundOcc-nuScenes while using only ~41M parameters in the occupancy head.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Embedding camera geometry into every attention and fusion step produces more accurate 3D semantic occupancy from multi-view images.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e8b6ff8d53bb1bed7734b93399bca0a617ffe01f86f2c8d77c478d741eff959f"},"source":{"id":"2605.16911","kind":"arxiv","version":1},"verdict":{"id":"9c0d051f-2d8c-4af8-b2d0-a43c1ad2c652","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T21:15:49.584724Z","strongest_claim":"By embedding geometric tokens throughout the pipeline with Projection-Aware Deformable Attention that projects 3D offsets and uses the projection Jacobian as an additive bias, plus a view-quality semantic gate and sequential coarse-to-fine gated fusion, VGGT-Occ achieves 33.00% IoU and 21.08% mIoU on SurroundOcc-nuScenes, outperforming prior methods with only ~41M trainable parameters in the occupancy head.","one_line_summary":"VGGT-Occ embeds geometric tokens via PA-DA and uses sequential coarse-to-fine gated fusion to reach 33.00% IoU and 21.08% mIoU on SurroundOcc-nuScenes while using only ~41M parameters in the occupancy head.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That projecting 3D offsets back to image planes and adding the projection Jacobian as a bias term will reliably suppress unreliable observations, and that the view-quality semantic gate will correctly enforce cross-view consistency without introducing new errors in feature integration.","pith_extraction_headline":"Embedding camera geometry into every attention and fusion step produces more accurate 3D semantic occupancy from multi-view images."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.16911/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T21:31:19.170645Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T21:20:56.520925Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T20:52:03.170953Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T18:41:56.269196Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T18:33:26.349114Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"f6d800d5d255a8a8cb183bc2dfba2463870158b52fc56f3cb192717725b663d0"},"references":{"count":50,"sample":[{"doi":"","year":2018,"title":"Maxim Berman, Amal Rannen Triki, and Matthew B. Blaschko. The Lovász-softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks. InCVPR, 2018","work_id":"38873ece-16b2-4c6e-9da8-e3065f2e5363","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Lang, Sourabh V ora, Venice Erin Liong, Qiang Xu, Anush Krishnan, Yu Pan, Giancarlo Baldan, and Oscar Beijbom","work_id":"7b09ba53-a1c3-41b7-a620-141ab87e3c71","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"MonoScene: Monocular 3D semantic scene completion","work_id":"81f3e606-939f-490a-ba91-292d3a456e1c","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Gauss- Render: Learning 3D occupancy with Gaussian rendering","work_id":"d1420371-0c70-4237-bdfe-45c1688aaea2","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Compact 3d gaussian splatting for dense visual slam","work_id":"d8274008-86ea-42d5-93fa-b5f806a9224a","ref_index":5,"cited_arxiv_id":"2403.11247","is_internal_anchor":true}],"resolved_work":50,"snapshot_sha256":"8ea1c05400a4573a82b32347be1afaf18fcd8cd5c82a299b4df637dc74eddc1d","internal_anchors":2},"formal_canon":{"evidence_count":2,"snapshot_sha256":"13d35d86d73e343896cc2b9629866c02b378f1cf9e2bc7cf100694880175c07f"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.16911","created_at":"2026-05-20T00:03:29.790104+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.16911v1","created_at":"2026-05-20T00:03:29.790104+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.16911","created_at":"2026-05-20T00:03:29.790104+00:00"},{"alias_kind":"pith_short_12","alias_value":"3BCWRAFNCLVN","created_at":"2026-05-20T00:03:29.790104+00:00"},{"alias_kind":"pith_short_16","alias_value":"3BCWRAFNCLVNTQV5","created_at":"2026-05-20T00:03:29.790104+00:00"},{"alias_kind":"pith_short_8","alias_value":"3BCWRAFN","created_at":"2026-05-20T00:03:29.790104+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3BCWRAFNCLVNTQV55HKVPBOZPC","json":"https://pith.science/pith/3BCWRAFNCLVNTQV55HKVPBOZPC.json","graph_json":"https://pith.science/api/pith-number/3BCWRAFNCLVNTQV55HKVPBOZPC/graph.json","events_json":"https://pith.science/api/pith-number/3BCWRAFNCLVNTQV55HKVPBOZPC/events.json","paper":"https://pith.science/paper/3BCWRAFN"},"agent_actions":{"view_html":"https://pith.science/pith/3BCWRAFNCLVNTQV55HKVPBOZPC","download_json":"https://pith.science/pith/3BCWRAFNCLVNTQV55HKVPBOZPC.json","view_paper":"https://pith.science/paper/3BCWRAFN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.16911&json=true","fetch_graph":"https://pith.science/api/pith-number/3BCWRAFNCLVNTQV55HKVPBOZPC/graph.json","fetch_events":"https://pith.science/api/pith-number/3BCWRAFNCLVNTQV55HKVPBOZPC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3BCWRAFNCLVNTQV55HKVPBOZPC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3BCWRAFNCLVNTQV55HKVPBOZPC/action/storage_attestation","attest_author":"https://pith.science/pith/3BCWRAFNCLVNTQV55HKVPBOZPC/action/author_attestation","sign_citation":"https://pith.science/pith/3BCWRAFNCLVNTQV55HKVPBOZPC/action/citation_signature","submit_replication":"https://pith.science/pith/3BCWRAFNCLVNTQV55HKVPBOZPC/action/replication_record"}},"created_at":"2026-05-20T00:03:29.790104+00:00","updated_at":"2026-05-20T00:03:29.790104+00:00"}