{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:YF43JYTFFZ5PSPDJQ6JOEBSPU3","short_pith_number":"pith:YF43JYTF","schema_version":"1.0","canonical_sha256":"c179b4e2652e7af93c698792e2064fa6ffc50d628432cf6386007910d9bb08fc","source":{"kind":"arxiv","id":"2602.00807","version":2},"attestation_state":"computed","paper":{"title":"Any3D-VLA: Enhancing VLA Robustness via Diverse Point Clouds","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Any3D-VLA improves VLA models by unifying simulator sensor and estimated point clouds into domain-agnostic 3D representations fused with 2D inputs.","cross_cats":["cs.RO"],"primary_cat":"cs.CV","authors_text":"Hengshuang Zhao, He Wang, Mi Yan, Shengliang Deng, Xianzhe Fan, Xiaoyang Wu, Yujia Zhang, Yuxiang Lu, Zhizheng Zhang, Zhuoling Li","submitted_at":"2026-01-31T16:34:52Z","abstract_excerpt":"Existing Vision-Language-Action (VLA) models typically take 2D images as visual input, which limits their spatial understanding in complex scenes. How can we incorporate 3D information to enhance VLA capabilities? We conduct a pilot study across different observation spaces and visual representations. The results show that explicitly lifting visual input into point clouds yields representations that better complement their corresponding 2D representations. To address the challenges of (1) scarce 3D data and (2) the domain gap induced by cross-environment differences and depth-scale biases, we "},"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":true},"canonical_record":{"source":{"id":"2602.00807","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-01-31T16:34:52Z","cross_cats_sorted":["cs.RO"],"title_canon_sha256":"07289c26013c377c2ef760c36d36984661f5ae6cfb6ecec8993110049672044e","abstract_canon_sha256":"503843f2c1d8c4bbce5ebae2a3b5080d6b636f76e7a0cd517fa84114dd7acec7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:00.121745Z","signature_b64":"Bu+fyw9EK2BJ3ddZsoo6yLoz/VlZ4k/MRw1zSg3YFVwl82p2fcLjfyw72UzlsxqNNZhgtzzLfuNM70NDn2VnCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c179b4e2652e7af93c698792e2064fa6ffc50d628432cf6386007910d9bb08fc","last_reissued_at":"2026-05-17T23:39:00.120954Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:00.120954Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Any3D-VLA: Enhancing VLA Robustness via Diverse Point Clouds","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Any3D-VLA improves VLA models by unifying simulator sensor and estimated point clouds into domain-agnostic 3D representations fused with 2D inputs.","cross_cats":["cs.RO"],"primary_cat":"cs.CV","authors_text":"Hengshuang Zhao, He Wang, Mi Yan, Shengliang Deng, Xianzhe Fan, Xiaoyang Wu, Yujia Zhang, Yuxiang Lu, Zhizheng Zhang, Zhuoling Li","submitted_at":"2026-01-31T16:34:52Z","abstract_excerpt":"Existing Vision-Language-Action (VLA) models typically take 2D images as visual input, which limits their spatial understanding in complex scenes. How can we incorporate 3D information to enhance VLA capabilities? We conduct a pilot study across different observation spaces and visual representations. The results show that explicitly lifting visual input into point clouds yields representations that better complement their corresponding 2D representations. To address the challenges of (1) scarce 3D data and (2) the domain gap induced by cross-environment differences and depth-scale biases, we "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Any3D-VLA unifies the simulator, sensor, and model-estimated point clouds within a training pipeline, constructs diverse inputs, and learns domain-agnostic 3D representations that are fused with the corresponding 2D representations, improving performance and mitigating the domain gap.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That explicitly lifting visual input into point clouds yields representations that better complement their corresponding 2D representations and that unifying across simulator, sensor, and estimated sources can close the domain gap without introducing new biases or performance drops.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Any3D-VLA unifies simulator, sensor, and estimated point clouds into domain-agnostic 3D features fused with 2D inputs to improve VLA robustness and reduce domain gaps.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Any3D-VLA improves VLA models by unifying simulator sensor and estimated point clouds into domain-agnostic 3D representations fused with 2D inputs.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"548916da0ce963c2ff8cb55f3e799eebe5347c2f1d7d3f4735d257eeb0587397"},"source":{"id":"2602.00807","kind":"arxiv","version":2},"verdict":{"id":"a934a382-04a0-4e2e-9920-56d92706cf6e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T08:43:09.630824Z","strongest_claim":"Any3D-VLA unifies the simulator, sensor, and model-estimated point clouds within a training pipeline, constructs diverse inputs, and learns domain-agnostic 3D representations that are fused with the corresponding 2D representations, improving performance and mitigating the domain gap.","one_line_summary":"Any3D-VLA unifies simulator, sensor, and estimated point clouds into domain-agnostic 3D features fused with 2D inputs to improve VLA robustness and reduce domain gaps.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That explicitly lifting visual input into point clouds yields representations that better complement their corresponding 2D representations and that unifying across simulator, sensor, and estimated sources can close the domain gap without introducing new biases or performance drops.","pith_extraction_headline":"Any3D-VLA improves VLA models by unifying simulator sensor and estimated point clouds into domain-agnostic 3D representations fused with 2D inputs."},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"cc4d0d08425e11f5c068dcfe6bdfb7a6cd789a2989801172125b75d82d10894d"},"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":"2602.00807","created_at":"2026-05-17T23:39:00.121094+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.00807v2","created_at":"2026-05-17T23:39:00.121094+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.00807","created_at":"2026-05-17T23:39:00.121094+00:00"},{"alias_kind":"pith_short_12","alias_value":"YF43JYTFFZ5P","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"YF43JYTFFZ5PSPDJ","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"YF43JYTF","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"2603.19199","citing_title":"FASTER: Rethinking Real-Time Flow VLAs","ref_index":19,"is_internal_anchor":true},{"citing_arxiv_id":"2605.20752","citing_title":"GaussianDream: A Feed-Forward 3D Gaussian World Model for Robotic Manipulation","ref_index":48,"is_internal_anchor":true},{"citing_arxiv_id":"2603.19199","citing_title":"FASTER: Rethinking Real-Time Flow VLAs","ref_index":19,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YF43JYTFFZ5PSPDJQ6JOEBSPU3","json":"https://pith.science/pith/YF43JYTFFZ5PSPDJQ6JOEBSPU3.json","graph_json":"https://pith.science/api/pith-number/YF43JYTFFZ5PSPDJQ6JOEBSPU3/graph.json","events_json":"https://pith.science/api/pith-number/YF43JYTFFZ5PSPDJQ6JOEBSPU3/events.json","paper":"https://pith.science/paper/YF43JYTF"},"agent_actions":{"view_html":"https://pith.science/pith/YF43JYTFFZ5PSPDJQ6JOEBSPU3","download_json":"https://pith.science/pith/YF43JYTFFZ5PSPDJQ6JOEBSPU3.json","view_paper":"https://pith.science/paper/YF43JYTF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.00807&json=true","fetch_graph":"https://pith.science/api/pith-number/YF43JYTFFZ5PSPDJQ6JOEBSPU3/graph.json","fetch_events":"https://pith.science/api/pith-number/YF43JYTFFZ5PSPDJQ6JOEBSPU3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YF43JYTFFZ5PSPDJQ6JOEBSPU3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YF43JYTFFZ5PSPDJQ6JOEBSPU3/action/storage_attestation","attest_author":"https://pith.science/pith/YF43JYTFFZ5PSPDJQ6JOEBSPU3/action/author_attestation","sign_citation":"https://pith.science/pith/YF43JYTFFZ5PSPDJQ6JOEBSPU3/action/citation_signature","submit_replication":"https://pith.science/pith/YF43JYTFFZ5PSPDJQ6JOEBSPU3/action/replication_record"}},"created_at":"2026-05-17T23:39:00.121094+00:00","updated_at":"2026-05-17T23:39:00.121094+00:00"}