{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:X2AUL54G4UDRDQJVYMM5RHO2OD","short_pith_number":"pith:X2AUL54G","schema_version":"1.0","canonical_sha256":"be8145f786e50711c135c319d89dda70daf887886a1ceaff5c91a55435267620","source":{"kind":"arxiv","id":"2506.15684","version":1},"attestation_state":"computed","paper":{"title":"Nabla-R2D3: Effective and Efficient 3D Diffusion Alignment with 2D Rewards","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.GR","authors_text":"Dinghuai Zhang, Kui Jia, Qingming Liu, Zhen Liu","submitted_at":"2025-06-18T17:59:59Z","abstract_excerpt":"Generating high-quality and photorealistic 3D assets remains a longstanding challenge in 3D vision and computer graphics. Although state-of-the-art generative models, such as diffusion models, have made significant progress in 3D generation, they often fall short of human-designed content due to limited ability to follow instructions, align with human preferences, or produce realistic textures, geometries, and physical attributes. In this paper, we introduce Nabla-R2D3, a highly effective and sample-efficient reinforcement learning alignment framework for 3D-native diffusion models using 2D re"},"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":"2506.15684","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.GR","submitted_at":"2025-06-18T17:59:59Z","cross_cats_sorted":["cs.CV","cs.LG"],"title_canon_sha256":"d4011af5064fcc9f6e616586eb15f469e22c22d73e81ae177e221883da30a6ff","abstract_canon_sha256":"e5b558f0f7fb2b33716753e66c29a1194606ed39d4c53f185cb535a00ab5f744"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:23:48.490802Z","signature_b64":"K7sMS6IClxsz0a5+6Oeh4tJw0MyvaSgsWJGcYR8q+6EHcYxPN8+NDL1Wc31m/Cv4i7Xlc31b0Ftw02/hJu1NBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"be8145f786e50711c135c319d89dda70daf887886a1ceaff5c91a55435267620","last_reissued_at":"2026-07-05T11:23:48.488833Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:23:48.488833Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Nabla-R2D3: Effective and Efficient 3D Diffusion Alignment with 2D Rewards","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.CV","cs.LG"],"primary_cat":"cs.GR","authors_text":"Dinghuai Zhang, Kui Jia, Qingming Liu, Zhen Liu","submitted_at":"2025-06-18T17:59:59Z","abstract_excerpt":"Generating high-quality and photorealistic 3D assets remains a longstanding challenge in 3D vision and computer graphics. Although state-of-the-art generative models, such as diffusion models, have made significant progress in 3D generation, they often fall short of human-designed content due to limited ability to follow instructions, align with human preferences, or produce realistic textures, geometries, and physical attributes. In this paper, we introduce Nabla-R2D3, a highly effective and sample-efficient reinforcement learning alignment framework for 3D-native diffusion models using 2D re"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.15684","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/2506.15684/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":"2506.15684","created_at":"2026-07-05T11:23:48.488898+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.15684v1","created_at":"2026-07-05T11:23:48.488898+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.15684","created_at":"2026-07-05T11:23:48.488898+00:00"},{"alias_kind":"pith_short_12","alias_value":"X2AUL54G4UDR","created_at":"2026-07-05T11:23:48.488898+00:00"},{"alias_kind":"pith_short_16","alias_value":"X2AUL54G4UDRDQJV","created_at":"2026-07-05T11:23:48.488898+00:00"},{"alias_kind":"pith_short_8","alias_value":"X2AUL54G","created_at":"2026-07-05T11:23:48.488898+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27305","citing_title":"Sculpting NeRF Geometry: Human-Preference Fine-Tuning of a 3D-Aware Face GAN","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18451","citing_title":"A Cross-Model VLM-Judge Protocol for Single-Image 3D Mesh Quality (and Why Cheap Proxies Fall Short)","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09711","citing_title":"Proxy Reward Internalization and Mechanistic Exploitation: A Learned Precursor to Reward Hacking and Its Generalization","ref_index":296,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13602","citing_title":"Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges","ref_index":176,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/X2AUL54G4UDRDQJVYMM5RHO2OD","json":"https://pith.science/pith/X2AUL54G4UDRDQJVYMM5RHO2OD.json","graph_json":"https://pith.science/api/pith-number/X2AUL54G4UDRDQJVYMM5RHO2OD/graph.json","events_json":"https://pith.science/api/pith-number/X2AUL54G4UDRDQJVYMM5RHO2OD/events.json","paper":"https://pith.science/paper/X2AUL54G"},"agent_actions":{"view_html":"https://pith.science/pith/X2AUL54G4UDRDQJVYMM5RHO2OD","download_json":"https://pith.science/pith/X2AUL54G4UDRDQJVYMM5RHO2OD.json","view_paper":"https://pith.science/paper/X2AUL54G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.15684&json=true","fetch_graph":"https://pith.science/api/pith-number/X2AUL54G4UDRDQJVYMM5RHO2OD/graph.json","fetch_events":"https://pith.science/api/pith-number/X2AUL54G4UDRDQJVYMM5RHO2OD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/X2AUL54G4UDRDQJVYMM5RHO2OD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/X2AUL54G4UDRDQJVYMM5RHO2OD/action/storage_attestation","attest_author":"https://pith.science/pith/X2AUL54G4UDRDQJVYMM5RHO2OD/action/author_attestation","sign_citation":"https://pith.science/pith/X2AUL54G4UDRDQJVYMM5RHO2OD/action/citation_signature","submit_replication":"https://pith.science/pith/X2AUL54G4UDRDQJVYMM5RHO2OD/action/replication_record"}},"created_at":"2026-07-05T11:23:48.488898+00:00","updated_at":"2026-07-05T11:23:48.488898+00:00"}