{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:OC6VL47P3IO4BZ3NB2GNTKB5VA","short_pith_number":"pith:OC6VL47P","schema_version":"1.0","canonical_sha256":"70bd55f3efda1dc0e76d0e8cd9a83da80c88db7f484a2a003993155d3b4928ff","source":{"kind":"arxiv","id":"2607.02291","version":1},"attestation_state":"computed","paper":{"title":"Optimizing Visual Generative Models via Distribution-wise Rewards","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Fuli Feng, Han Hu, Leigang Qu, Mengde Xu, Ruihang Li, Shuyang Gu, Wenjie Wang","submitted_at":"2026-07-02T15:08:56Z","abstract_excerpt":"Conventional reinforcement learning strategies for visual generation typically employ sample-wise reward functions, yet this practice frequently results in reward hacking that degrades image diversity and introduces visual anomalies. To address these limitations, we present a novel framework that finetunes generative models using distribution-wise rewards, ensuring better alignment with real-world data distributions. Unlike rewards that evaluate samples individually, distribution-wise reward accounts for the data distribution of the samples, mitigating the mode collapse problem that occurs whe"},"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":"2607.02291","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-02T15:08:56Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"d7e2382728ea5e215723108db16885db5a1c39ce04926e214e3c143906af238a","abstract_canon_sha256":"b425edd6959e0a13ce03bf44c6c64424f5765c5a5c825f9f6f6fe351d7c2c633"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-03T01:17:47.671523Z","signature_b64":"ZU3qZxtCOxI9XbY9xftmB9m8JBwzQbC9w2o8scCbg0cnPdhXXiQOM1qUT9L7z+U6yP7htE0gSR7AmxWBXhNcAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"70bd55f3efda1dc0e76d0e8cd9a83da80c88db7f484a2a003993155d3b4928ff","last_reissued_at":"2026-07-03T01:17:47.671098Z","signature_status":"signed_v1","first_computed_at":"2026-07-03T01:17:47.671098Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimizing Visual Generative Models via Distribution-wise Rewards","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Fuli Feng, Han Hu, Leigang Qu, Mengde Xu, Ruihang Li, Shuyang Gu, Wenjie Wang","submitted_at":"2026-07-02T15:08:56Z","abstract_excerpt":"Conventional reinforcement learning strategies for visual generation typically employ sample-wise reward functions, yet this practice frequently results in reward hacking that degrades image diversity and introduces visual anomalies. To address these limitations, we present a novel framework that finetunes generative models using distribution-wise rewards, ensuring better alignment with real-world data distributions. Unlike rewards that evaluate samples individually, distribution-wise reward accounts for the data distribution of the samples, mitigating the mode collapse problem that occurs whe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.02291","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/2607.02291/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":"2607.02291","created_at":"2026-07-03T01:17:47.671156+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.02291v1","created_at":"2026-07-03T01:17:47.671156+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.02291","created_at":"2026-07-03T01:17:47.671156+00:00"},{"alias_kind":"pith_short_12","alias_value":"OC6VL47P3IO4","created_at":"2026-07-03T01:17:47.671156+00:00"},{"alias_kind":"pith_short_16","alias_value":"OC6VL47P3IO4BZ3N","created_at":"2026-07-03T01:17:47.671156+00:00"},{"alias_kind":"pith_short_8","alias_value":"OC6VL47P","created_at":"2026-07-03T01:17:47.671156+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/OC6VL47P3IO4BZ3NB2GNTKB5VA","json":"https://pith.science/pith/OC6VL47P3IO4BZ3NB2GNTKB5VA.json","graph_json":"https://pith.science/api/pith-number/OC6VL47P3IO4BZ3NB2GNTKB5VA/graph.json","events_json":"https://pith.science/api/pith-number/OC6VL47P3IO4BZ3NB2GNTKB5VA/events.json","paper":"https://pith.science/paper/OC6VL47P"},"agent_actions":{"view_html":"https://pith.science/pith/OC6VL47P3IO4BZ3NB2GNTKB5VA","download_json":"https://pith.science/pith/OC6VL47P3IO4BZ3NB2GNTKB5VA.json","view_paper":"https://pith.science/paper/OC6VL47P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.02291&json=true","fetch_graph":"https://pith.science/api/pith-number/OC6VL47P3IO4BZ3NB2GNTKB5VA/graph.json","fetch_events":"https://pith.science/api/pith-number/OC6VL47P3IO4BZ3NB2GNTKB5VA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OC6VL47P3IO4BZ3NB2GNTKB5VA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OC6VL47P3IO4BZ3NB2GNTKB5VA/action/storage_attestation","attest_author":"https://pith.science/pith/OC6VL47P3IO4BZ3NB2GNTKB5VA/action/author_attestation","sign_citation":"https://pith.science/pith/OC6VL47P3IO4BZ3NB2GNTKB5VA/action/citation_signature","submit_replication":"https://pith.science/pith/OC6VL47P3IO4BZ3NB2GNTKB5VA/action/replication_record"}},"created_at":"2026-07-03T01:17:47.671156+00:00","updated_at":"2026-07-03T01:17:47.671156+00:00"}