{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:ZBWNMH3WD5TXFTGD6MX7FLLH4I","short_pith_number":"pith:ZBWNMH3W","schema_version":"1.0","canonical_sha256":"c86cd61f761f6772ccc3f32ff2ad67e22003ffbeac42ef32b66cea0ae745efac","source":{"kind":"arxiv","id":"2404.04465","version":2},"attestation_state":"computed","paper":{"title":"Aligning Diffusion Models by Optimizing Human Utility","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Akash Gokul, Kazuki Kozuka, Konstantinos Kallidromitis, Shufan Li, Yusuke Kato","submitted_at":"2024-04-06T01:23:23Z","abstract_excerpt":"We present Diffusion-KTO, a novel approach for aligning text-to-image diffusion models by formulating the alignment objective as the maximization of expected human utility. Since this objective applies to each generation independently, Diffusion-KTO does not require collecting costly pairwise preference data nor training a complex reward model. Instead, our objective requires simple per-image binary feedback signals, e.g. likes or dislikes, which are abundantly available. After fine-tuning using Diffusion-KTO, text-to-image diffusion models exhibit superior performance compared to existing tec"},"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":"2404.04465","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-04-06T01:23:23Z","cross_cats_sorted":[],"title_canon_sha256":"333f4bba01315da5c290bb6417f71bdb2db3f3b23fb9f6de825f106ed8979ffc","abstract_canon_sha256":"2b69ad0679448e15c3f8edb7833d0559dc8cc545fba36a84bdd62f4711809107"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:19:27.787066Z","signature_b64":"28WN9kJcEYBgNfhLd1E/PlfStwqa2nAwbBGfOVq+RQrxc2f88vGeOP3FKsHJg2eQkRfGaWsV0+x8wx6aaXaJCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c86cd61f761f6772ccc3f32ff2ad67e22003ffbeac42ef32b66cea0ae745efac","last_reissued_at":"2026-07-05T09:19:27.786586Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:19:27.786586Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Aligning Diffusion Models by Optimizing Human Utility","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Akash Gokul, Kazuki Kozuka, Konstantinos Kallidromitis, Shufan Li, Yusuke Kato","submitted_at":"2024-04-06T01:23:23Z","abstract_excerpt":"We present Diffusion-KTO, a novel approach for aligning text-to-image diffusion models by formulating the alignment objective as the maximization of expected human utility. Since this objective applies to each generation independently, Diffusion-KTO does not require collecting costly pairwise preference data nor training a complex reward model. Instead, our objective requires simple per-image binary feedback signals, e.g. likes or dislikes, which are abundantly available. After fine-tuning using Diffusion-KTO, text-to-image diffusion models exhibit superior performance compared to existing tec"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.04465","kind":"arxiv","version":2},"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/2404.04465/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":"2404.04465","created_at":"2026-07-05T09:19:27.786642+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.04465v2","created_at":"2026-07-05T09:19:27.786642+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.04465","created_at":"2026-07-05T09:19:27.786642+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZBWNMH3WD5TX","created_at":"2026-07-05T09:19:27.786642+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZBWNMH3WD5TXFTGD","created_at":"2026-07-05T09:19:27.786642+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZBWNMH3W","created_at":"2026-07-05T09:19:27.786642+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2406.03520","citing_title":"VideoPhy: Evaluating Physical Commonsense for Video Generation","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09433","citing_title":"Offline Preference Optimization for Rectified Flow with Noise-Tracked Pairs","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07427","citing_title":"Personalizing Text-to-Image Generation to Individual Taste","ref_index":31,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZBWNMH3WD5TXFTGD6MX7FLLH4I","json":"https://pith.science/pith/ZBWNMH3WD5TXFTGD6MX7FLLH4I.json","graph_json":"https://pith.science/api/pith-number/ZBWNMH3WD5TXFTGD6MX7FLLH4I/graph.json","events_json":"https://pith.science/api/pith-number/ZBWNMH3WD5TXFTGD6MX7FLLH4I/events.json","paper":"https://pith.science/paper/ZBWNMH3W"},"agent_actions":{"view_html":"https://pith.science/pith/ZBWNMH3WD5TXFTGD6MX7FLLH4I","download_json":"https://pith.science/pith/ZBWNMH3WD5TXFTGD6MX7FLLH4I.json","view_paper":"https://pith.science/paper/ZBWNMH3W","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.04465&json=true","fetch_graph":"https://pith.science/api/pith-number/ZBWNMH3WD5TXFTGD6MX7FLLH4I/graph.json","fetch_events":"https://pith.science/api/pith-number/ZBWNMH3WD5TXFTGD6MX7FLLH4I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZBWNMH3WD5TXFTGD6MX7FLLH4I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZBWNMH3WD5TXFTGD6MX7FLLH4I/action/storage_attestation","attest_author":"https://pith.science/pith/ZBWNMH3WD5TXFTGD6MX7FLLH4I/action/author_attestation","sign_citation":"https://pith.science/pith/ZBWNMH3WD5TXFTGD6MX7FLLH4I/action/citation_signature","submit_replication":"https://pith.science/pith/ZBWNMH3WD5TXFTGD6MX7FLLH4I/action/replication_record"}},"created_at":"2026-07-05T09:19:27.786642+00:00","updated_at":"2026-07-05T09:19:27.786642+00:00"}