{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:626O2D7NNATPKF2IB64FZJOPAL","short_pith_number":"pith:626O2D7N","schema_version":"1.0","canonical_sha256":"f6bced0fed6826f517480fb85ca5cf02f34a8afc20da10515871e1e7b9e146e9","source":{"kind":"arxiv","id":"2606.19162","version":1},"attestation_state":"computed","paper":{"title":"The Reward Was in Your Data All Along: Correcting Flow Matching with Discriminator-Guided RL","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Adriana Romero-Soriano, Felix Friedrich, Michal Drozdzal, Nicolas Beltran-Velez, Reyhane Askari-Hemmat, Xiaochuang Han, Zhang Xiaofeng","submitted_at":"2026-06-17T15:01:33Z","abstract_excerpt":"Score- and flow-matching models often rely on preference-based reinforcement learning for two purposes: aligning with subjective preferences and, surprisingly, recovering properties such as visual realism and coherent object structure that matching-based training is intended to learn from the data itself. We argue that this reflects a structural mismatch. Matching losses measure $\\ell_2$ regression error on the velocity or score field under training-time marginals, a proxy poorly aligned with the visual and semantic properties that determine sample quality at inference. Given a reward aligned "},"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":"2606.19162","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-17T15:01:33Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"4d6c23bda2a42136977cf9ff1bf19e4455560ea534ba170915c9b3e8fd98fa26","abstract_canon_sha256":"e3202d32e3dc93bc23b9e096b8f68d26a5a90ae03546a886f3e46638f8186bf6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:11:58.194306Z","signature_b64":"vVd/iTsYPIbvrj7nXs0kvWGU0VWHpelcOv2QyIQlB4gFIW1G3vDQAe01sdb1y47QEON9Oj12BKBPvfoPWhMNDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f6bced0fed6826f517480fb85ca5cf02f34a8afc20da10515871e1e7b9e146e9","last_reissued_at":"2026-06-19T16:11:58.193961Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:11:58.193961Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Reward Was in Your Data All Along: Correcting Flow Matching with Discriminator-Guided RL","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Adriana Romero-Soriano, Felix Friedrich, Michal Drozdzal, Nicolas Beltran-Velez, Reyhane Askari-Hemmat, Xiaochuang Han, Zhang Xiaofeng","submitted_at":"2026-06-17T15:01:33Z","abstract_excerpt":"Score- and flow-matching models often rely on preference-based reinforcement learning for two purposes: aligning with subjective preferences and, surprisingly, recovering properties such as visual realism and coherent object structure that matching-based training is intended to learn from the data itself. We argue that this reflects a structural mismatch. Matching losses measure $\\ell_2$ regression error on the velocity or score field under training-time marginals, a proxy poorly aligned with the visual and semantic properties that determine sample quality at inference. Given a reward aligned "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.19162","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/2606.19162/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":"2606.19162","created_at":"2026-06-19T16:11:58.194022+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.19162v1","created_at":"2026-06-19T16:11:58.194022+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.19162","created_at":"2026-06-19T16:11:58.194022+00:00"},{"alias_kind":"pith_short_12","alias_value":"626O2D7NNATP","created_at":"2026-06-19T16:11:58.194022+00:00"},{"alias_kind":"pith_short_16","alias_value":"626O2D7NNATPKF2I","created_at":"2026-06-19T16:11:58.194022+00:00"},{"alias_kind":"pith_short_8","alias_value":"626O2D7N","created_at":"2026-06-19T16:11:58.194022+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/626O2D7NNATPKF2IB64FZJOPAL","json":"https://pith.science/pith/626O2D7NNATPKF2IB64FZJOPAL.json","graph_json":"https://pith.science/api/pith-number/626O2D7NNATPKF2IB64FZJOPAL/graph.json","events_json":"https://pith.science/api/pith-number/626O2D7NNATPKF2IB64FZJOPAL/events.json","paper":"https://pith.science/paper/626O2D7N"},"agent_actions":{"view_html":"https://pith.science/pith/626O2D7NNATPKF2IB64FZJOPAL","download_json":"https://pith.science/pith/626O2D7NNATPKF2IB64FZJOPAL.json","view_paper":"https://pith.science/paper/626O2D7N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.19162&json=true","fetch_graph":"https://pith.science/api/pith-number/626O2D7NNATPKF2IB64FZJOPAL/graph.json","fetch_events":"https://pith.science/api/pith-number/626O2D7NNATPKF2IB64FZJOPAL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/626O2D7NNATPKF2IB64FZJOPAL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/626O2D7NNATPKF2IB64FZJOPAL/action/storage_attestation","attest_author":"https://pith.science/pith/626O2D7NNATPKF2IB64FZJOPAL/action/author_attestation","sign_citation":"https://pith.science/pith/626O2D7NNATPKF2IB64FZJOPAL/action/citation_signature","submit_replication":"https://pith.science/pith/626O2D7NNATPKF2IB64FZJOPAL/action/replication_record"}},"created_at":"2026-06-19T16:11:58.194022+00:00","updated_at":"2026-06-19T16:11:58.194022+00:00"}