{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:6UFK3XCFFOL2WLMAUBEUZQDP25","short_pith_number":"pith:6UFK3XCF","schema_version":"1.0","canonical_sha256":"f50aaddc452b97ab2d80a0494cc06fd7474e5e5cbe4b3c882c016beafa88ee6d","source":{"kind":"arxiv","id":"2602.07345","version":2},"attestation_state":"computed","paper":{"title":"Optimizing Few-Step Generation with Adaptive Matching Distillation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Bojun Chen, Lichen Bai, Shitong Shao, Shuo Chen, Shuo Yang, Wenliang Zhong, Zeke Xie, Zikai Zhou","submitted_at":"2026-02-07T04:00:20Z","abstract_excerpt":"Distribution Matching Distillation (DMD) is a powerful acceleration paradigm, yet its stability is often compromised in Forbidden Zone, regions where the real teacher provides unreliable guidance while the fake teacher exerts insufficient repulsive force. In this work, we propose a unified optimization framework that reinterprets prior art as implicit strategies to avoid these corrupted regions. Based on this insight, we introduce Adaptive Matching Distillation (AMD), a self-correcting mechanism that utilizes reward proxies to explicitly detect and escape Forbidden Zones. AMD dynamically prior"},"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":"2602.07345","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-02-07T04:00:20Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"01b575fe8c3ac14a791e4924f2767864ec1f0c842bd612e68766d5f0948a1e11","abstract_canon_sha256":"0ce1c83a30ddb1cf856dd13d88ceb3c428f8e9d6b81114067261e594f5856e47"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T02:07:19.920916Z","signature_b64":"WETign3UeXv+b3mKVg3HmHxUvwHNw9BPSQEPWJ7sI/uvOQFbExJvyoLAm6e1pcIOTbwxds0w4CMHAXnxCy7YDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f50aaddc452b97ab2d80a0494cc06fd7474e5e5cbe4b3c882c016beafa88ee6d","last_reissued_at":"2026-06-09T02:07:19.919870Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T02:07:19.919870Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimizing Few-Step Generation with Adaptive Matching Distillation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Bojun Chen, Lichen Bai, Shitong Shao, Shuo Chen, Shuo Yang, Wenliang Zhong, Zeke Xie, Zikai Zhou","submitted_at":"2026-02-07T04:00:20Z","abstract_excerpt":"Distribution Matching Distillation (DMD) is a powerful acceleration paradigm, yet its stability is often compromised in Forbidden Zone, regions where the real teacher provides unreliable guidance while the fake teacher exerts insufficient repulsive force. In this work, we propose a unified optimization framework that reinterprets prior art as implicit strategies to avoid these corrupted regions. Based on this insight, we introduce Adaptive Matching Distillation (AMD), a self-correcting mechanism that utilizes reward proxies to explicitly detect and escape Forbidden Zones. AMD dynamically prior"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.07345","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/2602.07345/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":"2602.07345","created_at":"2026-06-09T02:07:19.920013+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.07345v2","created_at":"2026-06-09T02:07:19.920013+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.07345","created_at":"2026-06-09T02:07:19.920013+00:00"},{"alias_kind":"pith_short_12","alias_value":"6UFK3XCFFOL2","created_at":"2026-06-09T02:07:19.920013+00:00"},{"alias_kind":"pith_short_16","alias_value":"6UFK3XCFFOL2WLMA","created_at":"2026-06-09T02:07:19.920013+00:00"},{"alias_kind":"pith_short_8","alias_value":"6UFK3XCF","created_at":"2026-06-09T02:07:19.920013+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":5,"sample":[{"citing_arxiv_id":"2605.14382","citing_title":"Delta Forcing: Trust Region Steering for Interactive Autoregressive Video Generation","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14382","citing_title":"Delta Forcing: Trust Region Steering for Interactive Autoregressive Video Generation","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14382","citing_title":"Delta Forcing: Trust Region Steering for Interactive Autoregressive Video Generation","ref_index":49,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07020","citing_title":"FlashMol: High-Quality Molecule Generation in as Few as Four Steps","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2604.15911","citing_title":"Efficient Video Diffusion Models: Advancements and Challenges","ref_index":211,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6UFK3XCFFOL2WLMAUBEUZQDP25","json":"https://pith.science/pith/6UFK3XCFFOL2WLMAUBEUZQDP25.json","graph_json":"https://pith.science/api/pith-number/6UFK3XCFFOL2WLMAUBEUZQDP25/graph.json","events_json":"https://pith.science/api/pith-number/6UFK3XCFFOL2WLMAUBEUZQDP25/events.json","paper":"https://pith.science/paper/6UFK3XCF"},"agent_actions":{"view_html":"https://pith.science/pith/6UFK3XCFFOL2WLMAUBEUZQDP25","download_json":"https://pith.science/pith/6UFK3XCFFOL2WLMAUBEUZQDP25.json","view_paper":"https://pith.science/paper/6UFK3XCF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.07345&json=true","fetch_graph":"https://pith.science/api/pith-number/6UFK3XCFFOL2WLMAUBEUZQDP25/graph.json","fetch_events":"https://pith.science/api/pith-number/6UFK3XCFFOL2WLMAUBEUZQDP25/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6UFK3XCFFOL2WLMAUBEUZQDP25/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6UFK3XCFFOL2WLMAUBEUZQDP25/action/storage_attestation","attest_author":"https://pith.science/pith/6UFK3XCFFOL2WLMAUBEUZQDP25/action/author_attestation","sign_citation":"https://pith.science/pith/6UFK3XCFFOL2WLMAUBEUZQDP25/action/citation_signature","submit_replication":"https://pith.science/pith/6UFK3XCFFOL2WLMAUBEUZQDP25/action/replication_record"}},"created_at":"2026-06-09T02:07:19.920013+00:00","updated_at":"2026-06-09T02:07:19.920013+00:00"}