{"paper":{"title":"FrameVGGT: Coherence-Preserving Memory for Bounded Streaming Geometry","license":"http://creativecommons.org/licenses/by/4.0/","headline":"FrameVGGT organizes each frame's KV contribution as a coherent segment summarized by key-space prototypes to bound memory while preserving multi-view geometric support in streaming VGGT.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Takeshi Oishi, Zhisong Xu","submitted_at":"2026-03-08T15:46:03Z","abstract_excerpt":"Streaming Visual Geometry Transformers such as StreamVGGT enable strong online 3D perception, but their KV-cache grows unbounded over long streams, limiting practical deployment. We study bounded-memory streaming geometry from the perspective of memory organization: unlike language modeling, where useful information can often be compressed at token level, geometry-driven inference relies on coherent and mutually compatible observations across views. Under fixed memory budgets, retaining history as isolated entries can progressively fragment the geometric context needed for stable long-horizon "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"FrameVGGT achieves favorable accuracy-memory trade-offs under bounded memory while maintaining more stable geometry over long streams.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That token-level retention fragments within-frame evidence and that summarizing segments with lightweight key-space prototypes preserves the redundant multi-view support needed for coherent geometric reasoning.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"FrameVGGT replaces token-level KV retention with frame-level segments and prototypes to bound memory while preserving geometric coherence in streaming VGGT.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"FrameVGGT organizes each frame's KV contribution as a coherent segment summarized by key-space prototypes to bound memory while preserving multi-view geometric support in streaming VGGT.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"940aeae480cac530028440f743e52d7db1fbd99b466bc554a886ab8d6d93ffe1"},"source":{"id":"2603.07690","kind":"arxiv","version":3},"verdict":{"id":"d6ea017a-ab35-4d2d-a978-005b4ef1762a","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T14:38:34.411159Z","strongest_claim":"FrameVGGT achieves favorable accuracy-memory trade-offs under bounded memory while maintaining more stable geometry over long streams.","one_line_summary":"FrameVGGT replaces token-level KV retention with frame-level segments and prototypes to bound memory while preserving geometric coherence in streaming VGGT.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That token-level retention fragments within-frame evidence and that summarizing segments with lightweight key-space prototypes preserves the redundant multi-view support needed for coherent geometric reasoning.","pith_extraction_headline":"FrameVGGT organizes each frame's KV contribution as a coherent segment summarized by key-space prototypes to bound memory while preserving multi-view geometric support in streaming VGGT."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.07690/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"f2db97f0649de34b3fc64622a7e590bd1af6371145185fe36c91be8afe041cbc"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}