{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5M2BSPTDL6TGUIMFS5XEI2GOUC","short_pith_number":"pith:5M2BSPTD","schema_version":"1.0","canonical_sha256":"eb34193e635fa66a2185976e4468cea09392f0056f53c537a59cd2b71608497d","source":{"kind":"arxiv","id":"2505.21893","version":3},"attestation_state":"computed","paper":{"title":"SIPO: Stabilized and Improved Preference Optimization for Aligning Diffusion Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hao Li, Junyan Wang, Mengping Yang, Xiaomeng Yang, Zhijian Zhou, Zhiyu Tan","submitted_at":"2025-05-28T02:11:56Z","abstract_excerpt":"Preference learning has garnered extensive attention as an effective technique for aligning diffusion models with human preferences in visual generation. However, existing alignment approaches such as Diffusion-DPO suffer from two fundamental challenges: training instability caused by high gradient variances at various timesteps and high parameter sensitivities, and off-policy bias arising from the discrepancy between the optimization data and the policy models' distribution. Our first contribution is a systematic analysis of diffusion trajectories across different timesteps, identifying that "},"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":"2505.21893","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-28T02:11:56Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"d82357d2f9c57c163f9e9fe87b16d7b730748ed4c1e73e210cac613170b2574e","abstract_canon_sha256":"cd0fc2afe74aa97ab83be67e8e57cc41d91bf83935d6afa9a39f8a1310f1be58"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:05:29.311276Z","signature_b64":"s+OJWAgLqj/i4SxN129QBmLN5/oOu51b+DVVPzKwMRhbXtxbqJmDml/4Ji3aHVqNQLHNBMmo31hMUeeJezPIBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eb34193e635fa66a2185976e4468cea09392f0056f53c537a59cd2b71608497d","last_reissued_at":"2026-05-20T00:05:29.310650Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:05:29.310650Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SIPO: Stabilized and Improved Preference Optimization for Aligning Diffusion Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hao Li, Junyan Wang, Mengping Yang, Xiaomeng Yang, Zhijian Zhou, Zhiyu Tan","submitted_at":"2025-05-28T02:11:56Z","abstract_excerpt":"Preference learning has garnered extensive attention as an effective technique for aligning diffusion models with human preferences in visual generation. However, existing alignment approaches such as Diffusion-DPO suffer from two fundamental challenges: training instability caused by high gradient variances at various timesteps and high parameter sensitivities, and off-policy bias arising from the discrepancy between the optimization data and the policy models' distribution. Our first contribution is a systematic analysis of diffusion trajectories across different timesteps, identifying that "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.21893","kind":"arxiv","version":3},"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/2505.21893/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":"2505.21893","created_at":"2026-05-20T00:05:29.310764+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.21893v3","created_at":"2026-05-20T00:05:29.310764+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.21893","created_at":"2026-05-20T00:05:29.310764+00:00"},{"alias_kind":"pith_short_12","alias_value":"5M2BSPTDL6TG","created_at":"2026-05-20T00:05:29.310764+00:00"},{"alias_kind":"pith_short_16","alias_value":"5M2BSPTDL6TGUIMF","created_at":"2026-05-20T00:05:29.310764+00:00"},{"alias_kind":"pith_short_8","alias_value":"5M2BSPTD","created_at":"2026-05-20T00:05:29.310764+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/5M2BSPTDL6TGUIMFS5XEI2GOUC","json":"https://pith.science/pith/5M2BSPTDL6TGUIMFS5XEI2GOUC.json","graph_json":"https://pith.science/api/pith-number/5M2BSPTDL6TGUIMFS5XEI2GOUC/graph.json","events_json":"https://pith.science/api/pith-number/5M2BSPTDL6TGUIMFS5XEI2GOUC/events.json","paper":"https://pith.science/paper/5M2BSPTD"},"agent_actions":{"view_html":"https://pith.science/pith/5M2BSPTDL6TGUIMFS5XEI2GOUC","download_json":"https://pith.science/pith/5M2BSPTDL6TGUIMFS5XEI2GOUC.json","view_paper":"https://pith.science/paper/5M2BSPTD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.21893&json=true","fetch_graph":"https://pith.science/api/pith-number/5M2BSPTDL6TGUIMFS5XEI2GOUC/graph.json","fetch_events":"https://pith.science/api/pith-number/5M2BSPTDL6TGUIMFS5XEI2GOUC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5M2BSPTDL6TGUIMFS5XEI2GOUC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5M2BSPTDL6TGUIMFS5XEI2GOUC/action/storage_attestation","attest_author":"https://pith.science/pith/5M2BSPTDL6TGUIMFS5XEI2GOUC/action/author_attestation","sign_citation":"https://pith.science/pith/5M2BSPTDL6TGUIMFS5XEI2GOUC/action/citation_signature","submit_replication":"https://pith.science/pith/5M2BSPTDL6TGUIMFS5XEI2GOUC/action/replication_record"}},"created_at":"2026-05-20T00:05:29.310764+00:00","updated_at":"2026-05-20T00:05:29.310764+00:00"}