{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:BPQPESIPOQMKPLOMXNTUROUX2R","short_pith_number":"pith:BPQPESIP","schema_version":"1.0","canonical_sha256":"0be0f2490f7418a7adccbb6748ba97d46ae1cbf1965ae56cca7463821471f4f3","source":{"kind":"arxiv","id":"2403.11027","version":2},"attestation_state":"computed","paper":{"title":"Reward Guided Latent Consistency Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Jiachen Li, Weixi Feng, Wenhu Chen, William Yang Wang","submitted_at":"2024-03-16T22:14:56Z","abstract_excerpt":"Latent Consistency Distillation (LCD) has emerged as a promising paradigm for efficient text-to-image synthesis. By distilling a latent consistency model (LCM) from a pre-trained teacher latent diffusion model (LDM), LCD facilitates the generation of high-fidelity images within merely 2 to 4 inference steps. However, the LCM's efficient inference is obtained at the cost of the sample quality. In this paper, we propose compensating the quality loss by aligning LCM's output with human preference during training. Specifically, we introduce Reward Guided LCD (RG-LCD), which integrates feedback fro"},"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":"2403.11027","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-03-16T22:14:56Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a7a133765b3ad44a25c5e3c08423ca2cb8c977d9a5e9c15ae2e66db72177d056","abstract_canon_sha256":"54c52c1a033a0f82fbb9b1deacc4026974f6ff9f203d66f0e2b9783acb998a62"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:17:12.286182Z","signature_b64":"WLbu6eGUdNRbXd3fF5CyuiUElNsj7deUcTczrPJIpbCjEpUcZjXAms6QNVtxDsY58spUH0CByX7ynrmaYsojDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0be0f2490f7418a7adccbb6748ba97d46ae1cbf1965ae56cca7463821471f4f3","last_reissued_at":"2026-07-05T09:17:12.285673Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:17:12.285673Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reward Guided Latent Consistency Distillation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Jiachen Li, Weixi Feng, Wenhu Chen, William Yang Wang","submitted_at":"2024-03-16T22:14:56Z","abstract_excerpt":"Latent Consistency Distillation (LCD) has emerged as a promising paradigm for efficient text-to-image synthesis. By distilling a latent consistency model (LCM) from a pre-trained teacher latent diffusion model (LDM), LCD facilitates the generation of high-fidelity images within merely 2 to 4 inference steps. However, the LCM's efficient inference is obtained at the cost of the sample quality. In this paper, we propose compensating the quality loss by aligning LCM's output with human preference during training. Specifically, we introduce Reward Guided LCD (RG-LCD), which integrates feedback fro"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.11027","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/2403.11027/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":"2403.11027","created_at":"2026-07-05T09:17:12.285730+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.11027v2","created_at":"2026-07-05T09:17:12.285730+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.11027","created_at":"2026-07-05T09:17:12.285730+00:00"},{"alias_kind":"pith_short_12","alias_value":"BPQPESIPOQMK","created_at":"2026-07-05T09:17:12.285730+00:00"},{"alias_kind":"pith_short_16","alias_value":"BPQPESIPOQMKPLOM","created_at":"2026-07-05T09:17:12.285730+00:00"},{"alias_kind":"pith_short_8","alias_value":"BPQPESIP","created_at":"2026-07-05T09:17:12.285730+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.32020","citing_title":"Cross-Space Distillation: Teaching One-Step Students with Modern Diffusion Teachers","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30414","citing_title":"Diffusion Fine-tuning with Rewarded Moment Matching Distillation","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21484","citing_title":"One-Step Distillation of Discrete Diffusion Image Generators via Fixed-Point Iteration","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06916","citing_title":"FP4 Explore, BF16 Train: Diffusion Reinforcement Learning via Efficient Rollout Scaling","ref_index":33,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BPQPESIPOQMKPLOMXNTUROUX2R","json":"https://pith.science/pith/BPQPESIPOQMKPLOMXNTUROUX2R.json","graph_json":"https://pith.science/api/pith-number/BPQPESIPOQMKPLOMXNTUROUX2R/graph.json","events_json":"https://pith.science/api/pith-number/BPQPESIPOQMKPLOMXNTUROUX2R/events.json","paper":"https://pith.science/paper/BPQPESIP"},"agent_actions":{"view_html":"https://pith.science/pith/BPQPESIPOQMKPLOMXNTUROUX2R","download_json":"https://pith.science/pith/BPQPESIPOQMKPLOMXNTUROUX2R.json","view_paper":"https://pith.science/paper/BPQPESIP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.11027&json=true","fetch_graph":"https://pith.science/api/pith-number/BPQPESIPOQMKPLOMXNTUROUX2R/graph.json","fetch_events":"https://pith.science/api/pith-number/BPQPESIPOQMKPLOMXNTUROUX2R/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BPQPESIPOQMKPLOMXNTUROUX2R/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BPQPESIPOQMKPLOMXNTUROUX2R/action/storage_attestation","attest_author":"https://pith.science/pith/BPQPESIPOQMKPLOMXNTUROUX2R/action/author_attestation","sign_citation":"https://pith.science/pith/BPQPESIPOQMKPLOMXNTUROUX2R/action/citation_signature","submit_replication":"https://pith.science/pith/BPQPESIPOQMKPLOMXNTUROUX2R/action/replication_record"}},"created_at":"2026-07-05T09:17:12.285730+00:00","updated_at":"2026-07-05T09:17:12.285730+00:00"}