{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EO63BI7RVRSQA7CUSM72GTAXP7","short_pith_number":"pith:EO63BI7R","schema_version":"1.0","canonical_sha256":"23bdb0a3f1ac65007c54933fa34c177fe2a6b6407b780c9ba03e84fdc32737e5","source":{"kind":"arxiv","id":"2605.26108","version":1},"attestation_state":"computed","paper":{"title":"Reinforcing Few-step Generators via Reward-Tilted Distribution Matching","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chi Zhang, Jun Zhang, Ruoyu Wang, Tianyu Pang, Xiangxin Zhou, Yushi Huang","submitted_at":"2026-05-25T17:59:21Z","abstract_excerpt":"Recent advances in few-step diffusion distillation have enabled efficient image generation, yet aligning these models with human preferences remains challenging. We propose Reward-Tilted Distribution Matching Distillation (RTDMD), a two-stage framework that unifies distribution matching distillation with reward-guided reinforcement learning for few-step flow generators. We show that minimizing the KL divergence to a reward-tilted teacher distribution naturally decomposes into a distribution matching term and a reward maximization term. In the first stage, we introduce Ambient-Consistent Distri"},"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":"2605.26108","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-25T17:59:21Z","cross_cats_sorted":[],"title_canon_sha256":"5ae12131a73adbcc70e1f2aa922f7b8dcc78f8d95a90f02b23a8fc51ebc21180","abstract_canon_sha256":"71ccc7f01fe7f8c5436e963a1569c05f61e011797a60149febffe0803286d98c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T03:05:24.494707Z","signature_b64":"l9VXVq7LVmypRv4u87MmfEn9ED+CR3BbmzBd5jGDaBRjgIO0ot14hw0/h61T8QmyGplXGuOhtRBwkiFDoI5vCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"23bdb0a3f1ac65007c54933fa34c177fe2a6b6407b780c9ba03e84fdc32737e5","last_reissued_at":"2026-05-26T03:05:24.493686Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T03:05:24.493686Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reinforcing Few-step Generators via Reward-Tilted Distribution Matching","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chi Zhang, Jun Zhang, Ruoyu Wang, Tianyu Pang, Xiangxin Zhou, Yushi Huang","submitted_at":"2026-05-25T17:59:21Z","abstract_excerpt":"Recent advances in few-step diffusion distillation have enabled efficient image generation, yet aligning these models with human preferences remains challenging. We propose Reward-Tilted Distribution Matching Distillation (RTDMD), a two-stage framework that unifies distribution matching distillation with reward-guided reinforcement learning for few-step flow generators. We show that minimizing the KL divergence to a reward-tilted teacher distribution naturally decomposes into a distribution matching term and a reward maximization term. In the first stage, we introduce Ambient-Consistent Distri"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.26108","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/2605.26108/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":"2605.26108","created_at":"2026-05-26T03:05:24.493833+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.26108v1","created_at":"2026-05-26T03:05:24.493833+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.26108","created_at":"2026-05-26T03:05:24.493833+00:00"},{"alias_kind":"pith_short_12","alias_value":"EO63BI7RVRSQ","created_at":"2026-05-26T03:05:24.493833+00:00"},{"alias_kind":"pith_short_16","alias_value":"EO63BI7RVRSQA7CU","created_at":"2026-05-26T03:05:24.493833+00:00"},{"alias_kind":"pith_short_8","alias_value":"EO63BI7R","created_at":"2026-05-26T03:05:24.493833+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/EO63BI7RVRSQA7CUSM72GTAXP7","json":"https://pith.science/pith/EO63BI7RVRSQA7CUSM72GTAXP7.json","graph_json":"https://pith.science/api/pith-number/EO63BI7RVRSQA7CUSM72GTAXP7/graph.json","events_json":"https://pith.science/api/pith-number/EO63BI7RVRSQA7CUSM72GTAXP7/events.json","paper":"https://pith.science/paper/EO63BI7R"},"agent_actions":{"view_html":"https://pith.science/pith/EO63BI7RVRSQA7CUSM72GTAXP7","download_json":"https://pith.science/pith/EO63BI7RVRSQA7CUSM72GTAXP7.json","view_paper":"https://pith.science/paper/EO63BI7R","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.26108&json=true","fetch_graph":"https://pith.science/api/pith-number/EO63BI7RVRSQA7CUSM72GTAXP7/graph.json","fetch_events":"https://pith.science/api/pith-number/EO63BI7RVRSQA7CUSM72GTAXP7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EO63BI7RVRSQA7CUSM72GTAXP7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EO63BI7RVRSQA7CUSM72GTAXP7/action/storage_attestation","attest_author":"https://pith.science/pith/EO63BI7RVRSQA7CUSM72GTAXP7/action/author_attestation","sign_citation":"https://pith.science/pith/EO63BI7RVRSQA7CUSM72GTAXP7/action/citation_signature","submit_replication":"https://pith.science/pith/EO63BI7RVRSQA7CUSM72GTAXP7/action/replication_record"}},"created_at":"2026-05-26T03:05:24.493833+00:00","updated_at":"2026-05-26T03:05:24.493833+00:00"}