{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:V76JANKCHRE6QXKKEXOKUCG6OD","short_pith_number":"pith:V76JANKC","schema_version":"1.0","canonical_sha256":"affc9035423c49e85d4a25dcaa08de70f49cae17946b95f660eb31334f04fbf9","source":{"kind":"arxiv","id":"2503.07511","version":1},"attestation_state":"computed","paper":{"title":"PointVLA: Injecting the 3D World into Vision-Language-Action Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.RO","authors_text":"Chengmeng Li, Feifei Feng, Junjie Wen, Yan Peng, Yaxin Peng, Yichen Zhu","submitted_at":"2025-03-10T16:32:41Z","abstract_excerpt":"Vision-Language-Action (VLA) models excel at robotic tasks by leveraging large-scale 2D vision-language pretraining, but their reliance on RGB images limits spatial reasoning critical for real-world interaction. Retraining these models with 3D data is computationally prohibitive, while discarding existing 2D datasets wastes valuable resources. To bridge this gap, we propose PointVLA, a framework that enhances pre-trained VLAs with point cloud inputs without requiring retraining. Our method freezes the vanilla action expert and injects 3D features via a lightweight modular block. To identify th"},"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":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2503.07511","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2025-03-10T16:32:41Z","cross_cats_sorted":["cs.CV","cs.LG"],"title_canon_sha256":"131b59cacd1324490e955904ce2459eafdcbb08bd3115b70ca484efa421c522c","abstract_canon_sha256":"5257025f6c190ac15a4b701b5c5a78cd349163a42947e2425f5415ed9548071d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:28:06.115323Z","signature_b64":"YQ32Fa9Tn448/hOFoOVMk2+OZPvGLGKdXcXKWRS1Dq+/b4JdI1Dn7olfCA6sljLFP5W4nftdMeySS9il7vXcCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"affc9035423c49e85d4a25dcaa08de70f49cae17946b95f660eb31334f04fbf9","last_reissued_at":"2026-07-05T10:28:06.114647Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:28:06.114647Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PointVLA: Injecting the 3D World into Vision-Language-Action Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.RO","authors_text":"Chengmeng Li, Feifei Feng, Junjie Wen, Yan Peng, Yaxin Peng, Yichen Zhu","submitted_at":"2025-03-10T16:32:41Z","abstract_excerpt":"Vision-Language-Action (VLA) models excel at robotic tasks by leveraging large-scale 2D vision-language pretraining, but their reliance on RGB images limits spatial reasoning critical for real-world interaction. Retraining these models with 3D data is computationally prohibitive, while discarding existing 2D datasets wastes valuable resources. To bridge this gap, we propose PointVLA, a framework that enhances pre-trained VLAs with point cloud inputs without requiring retraining. Our method freezes the vanilla action expert and injects 3D features via a lightweight modular block. To identify th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.07511","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2503.07511/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":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"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":"2503.07511","created_at":"2026-07-05T10:28:06.114721+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.07511v1","created_at":"2026-07-05T10:28:06.114721+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.07511","created_at":"2026-07-05T10:28:06.114721+00:00"},{"alias_kind":"pith_short_12","alias_value":"V76JANKCHRE6","created_at":"2026-07-05T10:28:06.114721+00:00"},{"alias_kind":"pith_short_16","alias_value":"V76JANKCHRE6QXKK","created_at":"2026-07-05T10:28:06.114721+00:00"},{"alias_kind":"pith_short_8","alias_value":"V76JANKC","created_at":"2026-07-05T10:28:06.114721+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":15,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.13394","citing_title":"GeoHAT: Geometry-Adaptive Hybrid Action Transformer for Mobile Manipulation","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12403","citing_title":"World Pilot: Steering Vision-Language-Action Models with World-Action Priors","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22671","citing_title":"From Abstraction to Instantiation: Learning Behavioral Representation for Vision-Language-Action Model","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24642","citing_title":"Understanding the Impact of Geometric Foundation Models on Vision-Language-Action Models","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25829","citing_title":"OASIS: Observation-Action Space Alignment via SE(3) Trajectory Prediction for Robotic Manipulation","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22671","citing_title":"From Abstraction to Instantiation: Learning Behavioral Representation for Vision-Language-Action Model","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2506.13456","citing_title":"Block-wise Adaptive Caching for Accelerating Diffusion Policy","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2508.13073","citing_title":"Large VLM-based Vision-Language-Action Models for Robotic Manipulation: A Survey","ref_index":161,"is_internal_anchor":false},{"citing_arxiv_id":"2512.23864","citing_title":"Learning to Feel the Future: DreamTacVLA for Contact-Rich Manipulation","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12162","citing_title":"X-Imitator: Spatial-Aware Imitation Learning via Bidirectional Action-Pose Interaction","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05126","citing_title":"ConsisVLA-4D: Advancing Spatiotemporal Consistency in Efficient 3D-Perception and 4D-Reasoning for Robotic Manipulation","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04834","citing_title":"E-VLA: Event-Augmented Vision-Language-Action Model for Dark and Blurred Scenes","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17880","citing_title":"ST-$\\pi$: Structured SpatioTemporal VLA for Robotic Manipulation","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15281","citing_title":"R3D: Revisiting 3D Policy Learning","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20834","citing_title":"PokeVLA: Empowering Pocket-Sized Vision-Language-Action Model with Comprehensive World Knowledge Guidance","ref_index":36,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/V76JANKCHRE6QXKKEXOKUCG6OD","json":"https://pith.science/pith/V76JANKCHRE6QXKKEXOKUCG6OD.json","graph_json":"https://pith.science/api/pith-number/V76JANKCHRE6QXKKEXOKUCG6OD/graph.json","events_json":"https://pith.science/api/pith-number/V76JANKCHRE6QXKKEXOKUCG6OD/events.json","paper":"https://pith.science/paper/V76JANKC"},"agent_actions":{"view_html":"https://pith.science/pith/V76JANKCHRE6QXKKEXOKUCG6OD","download_json":"https://pith.science/pith/V76JANKCHRE6QXKKEXOKUCG6OD.json","view_paper":"https://pith.science/paper/V76JANKC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.07511&json=true","fetch_graph":"https://pith.science/api/pith-number/V76JANKCHRE6QXKKEXOKUCG6OD/graph.json","fetch_events":"https://pith.science/api/pith-number/V76JANKCHRE6QXKKEXOKUCG6OD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V76JANKCHRE6QXKKEXOKUCG6OD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V76JANKCHRE6QXKKEXOKUCG6OD/action/storage_attestation","attest_author":"https://pith.science/pith/V76JANKCHRE6QXKKEXOKUCG6OD/action/author_attestation","sign_citation":"https://pith.science/pith/V76JANKCHRE6QXKKEXOKUCG6OD/action/citation_signature","submit_replication":"https://pith.science/pith/V76JANKCHRE6QXKKEXOKUCG6OD/action/replication_record"}},"created_at":"2026-07-05T10:28:06.114721+00:00","updated_at":"2026-07-05T10:28:06.114721+00:00"}