{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:IOVTADSAIQNRQIZOM5ICML5RTE","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"965112703ecf90a21ea7fa74bb4649f763107130cbece4ac12a9ddcc00b3251d","cross_cats_sorted":["cs.AI","cs.CV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-19T12:47:52Z","title_canon_sha256":"a9b9c9999d2b319856cd98f141b6c5316e3159c1ff90fcf1fc01e0069613e15f"},"schema_version":"1.0","source":{"id":"2604.17415","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.17415","created_at":"2026-06-02T01:03:47Z"},{"alias_kind":"arxiv_version","alias_value":"2604.17415v3","created_at":"2026-06-02T01:03:47Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.17415","created_at":"2026-06-02T01:03:47Z"},{"alias_kind":"pith_short_12","alias_value":"IOVTADSAIQNR","created_at":"2026-06-02T01:03:47Z"},{"alias_kind":"pith_short_16","alias_value":"IOVTADSAIQNRQIZO","created_at":"2026-06-02T01:03:47Z"},{"alias_kind":"pith_short_8","alias_value":"IOVTADSA","created_at":"2026-06-02T01:03:47Z"}],"graph_snapshots":[{"event_id":"sha256:5697b95e8df7243c1c71adecbafb634e3b8e09e0403bbf0b33d66f685b490d08","target":"graph","created_at":"2026-06-02T01:03:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Although existing methods are derived from different perspectives, we show that many can be written under a common framework, which we call reward score matching (RSM)."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the primary distinctions among existing methods reduce to the construction of the value-guidance estimator and the effective optimization strength across timesteps, without material loss of generality or overlooked auxiliary mechanisms."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Reward Score Matching unifies reward-based fine-tuning for flow and diffusion models by recasting alignment as score matching to a value-guided target."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Many reward-based fine-tuning methods for diffusion and flow models reduce to a single score-matching objective against a value-guided target."}],"snapshot_sha256":"4a6e37e38fb316c83f3835af855adce2bbccf2923c582855d1481dae5a07005b"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.17415/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reward-based fine-tuning steers a pretrained diffusion or flow-based generative model toward higher-reward samples while remaining close to the pretrained model. Although existing methods are derived from different perspectives, we show that many can be written under a common framework, which we call reward score matching (RSM). Under this view, alignment becomes score matching against a value-guided target, and the main differences across methods reduce to the construction of the value-guidance estimator and the effective optimization strength across timesteps. This unification clarifies the ","authors_text":"Jeongjae Lee, Jeongsol Kim, Jinho Chang, Jong Chul Ye","cross_cats":["cs.AI","cs.CV"],"headline":"Many reward-based fine-tuning methods for diffusion and flow models reduce to a single score-matching objective against a value-guided target.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-19T12:47:52Z","title":"Reward Score Matching: Unifying Reward-based Fine-tuning for Flow and Diffusion Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.17415","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-10T05:30:29.711746Z","id":"be8edd1a-49b9-46d3-ba94-a222182b8113","model_set":{"reader":"grok-4.3"},"one_line_summary":"Reward Score Matching unifies reward-based fine-tuning for flow and diffusion models by recasting alignment as score matching to a value-guided target.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Many reward-based fine-tuning methods for diffusion and flow models reduce to a single score-matching objective against a value-guided target.","strongest_claim":"Although existing methods are derived from different perspectives, we show that many can be written under a common framework, which we call reward score matching (RSM).","weakest_assumption":"That the primary distinctions among existing methods reduce to the construction of the value-guidance estimator and the effective optimization strength across timesteps, without material loss of generality or overlooked auxiliary mechanisms."}},"verdict_id":"be8edd1a-49b9-46d3-ba94-a222182b8113"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:008857f3884c6345f9ea4cb6142d6061174e62b4d0a095f89e429627ac8101b1","target":"record","created_at":"2026-06-02T01:03:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"965112703ecf90a21ea7fa74bb4649f763107130cbece4ac12a9ddcc00b3251d","cross_cats_sorted":["cs.AI","cs.CV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-04-19T12:47:52Z","title_canon_sha256":"a9b9c9999d2b319856cd98f141b6c5316e3159c1ff90fcf1fc01e0069613e15f"},"schema_version":"1.0","source":{"id":"2604.17415","kind":"arxiv","version":3}},"canonical_sha256":"43ab300e40441b18232e6750262fb19903d11d9b06c5390de10b5769fee83e8f","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"43ab300e40441b18232e6750262fb19903d11d9b06c5390de10b5769fee83e8f","first_computed_at":"2026-06-02T01:03:47.239586Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-02T01:03:47.239586Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"stWrP+g57acQKZrSwwom9zJ9sJUEl99Du3ZocIUXw5N3CFk70zaXftmv206rNfz+7dULLWpTV2bfZgmKU2VGAg==","signature_status":"signed_v1","signed_at":"2026-06-02T01:03:47.240129Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.17415","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:008857f3884c6345f9ea4cb6142d6061174e62b4d0a095f89e429627ac8101b1","sha256:5697b95e8df7243c1c71adecbafb634e3b8e09e0403bbf0b33d66f685b490d08"],"state_sha256":"e18244bf465a58b61c7b71acb6d4615eeb7f9c725b65c969416e393c76e7a187"}