{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:I6XN4PZFYTQUP6WKHNSL7LNUDI","short_pith_number":"pith:I6XN4PZF","schema_version":"1.0","canonical_sha256":"47aede3f25c4e147faca3b64bfadb41a2a40d0dba2e6ab9c7fbc7290f57b8073","source":{"kind":"arxiv","id":"2405.13637","version":6},"attestation_state":"computed","paper":{"title":"Curriculum Direct Preference Optimization for Diffusion and Consistency Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Florinel-Alin Croitoru, Mubarak Shah, Nicu Sebe, Radu Tudor Ionescu, Vlad Hondru","submitted_at":"2024-05-22T13:36:48Z","abstract_excerpt":"Direct Preference Optimization (DPO) has been proposed as an effective and efficient alternative to reinforcement learning from human feedback (RLHF). In this paper, we propose a novel and enhanced version of DPO based on curriculum learning for text-to-image generation. Our method is divided into two training stages. First, a ranking of the examples generated for each prompt is obtained by employing a reward model. Then, increasingly difficult pairs of examples are sampled and provided to a text-to-image generative (diffusion or consistency) model. Generated samples that are far apart in the "},"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":"2405.13637","kind":"arxiv","version":6},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-05-22T13:36:48Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"71c31d2bb090ea2a1a4c03498ae27ea47ce53f63bdab7fd1f9bbdec912915e91","abstract_canon_sha256":"1c68a80b3b0d7432f7e7aab85d8aaac0a514497043d249ce3dfbb6d172689520"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:00:37.130474Z","signature_b64":"gnAL8BfxUKM0i8+kLyO7oGcj2mdoShh3XGFr7cjmhuOcPzDJIWhib4T1iQsTjwXj/Z9ct4TPVk5TeVtZBA9dDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"47aede3f25c4e147faca3b64bfadb41a2a40d0dba2e6ab9c7fbc7290f57b8073","last_reissued_at":"2026-07-05T11:00:37.129977Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:00:37.129977Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Curriculum Direct Preference Optimization for Diffusion and Consistency Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Florinel-Alin Croitoru, Mubarak Shah, Nicu Sebe, Radu Tudor Ionescu, Vlad Hondru","submitted_at":"2024-05-22T13:36:48Z","abstract_excerpt":"Direct Preference Optimization (DPO) has been proposed as an effective and efficient alternative to reinforcement learning from human feedback (RLHF). In this paper, we propose a novel and enhanced version of DPO based on curriculum learning for text-to-image generation. Our method is divided into two training stages. First, a ranking of the examples generated for each prompt is obtained by employing a reward model. Then, increasingly difficult pairs of examples are sampled and provided to a text-to-image generative (diffusion or consistency) model. Generated samples that are far apart in the "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.13637","kind":"arxiv","version":6},"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/2405.13637/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":"2405.13637","created_at":"2026-07-05T11:00:37.130033+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.13637v6","created_at":"2026-07-05T11:00:37.130033+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.13637","created_at":"2026-07-05T11:00:37.130033+00:00"},{"alias_kind":"pith_short_12","alias_value":"I6XN4PZFYTQU","created_at":"2026-07-05T11:00:37.130033+00:00"},{"alias_kind":"pith_short_16","alias_value":"I6XN4PZFYTQUP6WK","created_at":"2026-07-05T11:00:37.130033+00:00"},{"alias_kind":"pith_short_8","alias_value":"I6XN4PZF","created_at":"2026-07-05T11:00:37.130033+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.09433","citing_title":"Offline Preference Optimization for Rectified Flow with Noise-Tracked Pairs","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/I6XN4PZFYTQUP6WKHNSL7LNUDI","json":"https://pith.science/pith/I6XN4PZFYTQUP6WKHNSL7LNUDI.json","graph_json":"https://pith.science/api/pith-number/I6XN4PZFYTQUP6WKHNSL7LNUDI/graph.json","events_json":"https://pith.science/api/pith-number/I6XN4PZFYTQUP6WKHNSL7LNUDI/events.json","paper":"https://pith.science/paper/I6XN4PZF"},"agent_actions":{"view_html":"https://pith.science/pith/I6XN4PZFYTQUP6WKHNSL7LNUDI","download_json":"https://pith.science/pith/I6XN4PZFYTQUP6WKHNSL7LNUDI.json","view_paper":"https://pith.science/paper/I6XN4PZF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.13637&json=true","fetch_graph":"https://pith.science/api/pith-number/I6XN4PZFYTQUP6WKHNSL7LNUDI/graph.json","fetch_events":"https://pith.science/api/pith-number/I6XN4PZFYTQUP6WKHNSL7LNUDI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/I6XN4PZFYTQUP6WKHNSL7LNUDI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/I6XN4PZFYTQUP6WKHNSL7LNUDI/action/storage_attestation","attest_author":"https://pith.science/pith/I6XN4PZFYTQUP6WKHNSL7LNUDI/action/author_attestation","sign_citation":"https://pith.science/pith/I6XN4PZFYTQUP6WKHNSL7LNUDI/action/citation_signature","submit_replication":"https://pith.science/pith/I6XN4PZFYTQUP6WKHNSL7LNUDI/action/replication_record"}},"created_at":"2026-07-05T11:00:37.130033+00:00","updated_at":"2026-07-05T11:00:37.130033+00:00"}