{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:GV6322X3BC3SHZN5PWXZTTEPX3","short_pith_number":"pith:GV6322X3","schema_version":"1.0","canonical_sha256":"357dbd6afb08b723e5bd7daf99cc8fbeec1c41a24b11212078c1049e54210dea","source":{"kind":"arxiv","id":"2405.05224","version":1},"attestation_state":"computed","paper":{"title":"Imagine Flash: Accelerating Emu Diffusion Models with Backward Distillation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Albert Pumarola, Ali Thabet, Artsiom Sanakoyeu, Edgar Sch\\\"onfeld, Jonas Kohler, Peter Vajda, Roshan Sumbaly","submitted_at":"2024-05-08T17:15:18Z","abstract_excerpt":"Diffusion models are a powerful generative framework, but come with expensive inference. Existing acceleration methods often compromise image quality or fail under complex conditioning when operating in an extremely low-step regime. In this work, we propose a novel distillation framework tailored to enable high-fidelity, diverse sample generation using just one to three steps. Our approach comprises three key components: (i) Backward Distillation, which mitigates training-inference discrepancies by calibrating the student on its own backward trajectory; (ii) Shifted Reconstruction Loss that dy"},"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":"2405.05224","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-05-08T17:15:18Z","cross_cats_sorted":[],"title_canon_sha256":"4d07a288ccc240a3cad589d0e4744f3a2cb87ba014302e796b21fa94285d9d08","abstract_canon_sha256":"1f0637c57c2e5d11325ecb5aa5347c214ce6bdae6fc350817184b66d6cd68c30"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:17:06.697947Z","signature_b64":"UdEQhQxaxu6quw11fFRu2dJovmPMDRpWqONEEBS6nzGnqvenYSSUckuZym6q+mnJHhO8so7HrFA2y5l7TyWNCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"357dbd6afb08b723e5bd7daf99cc8fbeec1c41a24b11212078c1049e54210dea","last_reissued_at":"2026-07-05T08:17:06.697466Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:17:06.697466Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Imagine Flash: Accelerating Emu Diffusion Models with Backward Distillation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Albert Pumarola, Ali Thabet, Artsiom Sanakoyeu, Edgar Sch\\\"onfeld, Jonas Kohler, Peter Vajda, Roshan Sumbaly","submitted_at":"2024-05-08T17:15:18Z","abstract_excerpt":"Diffusion models are a powerful generative framework, but come with expensive inference. Existing acceleration methods often compromise image quality or fail under complex conditioning when operating in an extremely low-step regime. In this work, we propose a novel distillation framework tailored to enable high-fidelity, diverse sample generation using just one to three steps. Our approach comprises three key components: (i) Backward Distillation, which mitigates training-inference discrepancies by calibrating the student on its own backward trajectory; (ii) Shifted Reconstruction Loss that dy"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.05224","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/2405.05224/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":"2405.05224","created_at":"2026-07-05T08:17:06.697532+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.05224v1","created_at":"2026-07-05T08:17:06.697532+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.05224","created_at":"2026-07-05T08:17:06.697532+00:00"},{"alias_kind":"pith_short_12","alias_value":"GV6322X3BC3S","created_at":"2026-07-05T08:17:06.697532+00:00"},{"alias_kind":"pith_short_16","alias_value":"GV6322X3BC3SHZN5","created_at":"2026-07-05T08:17:06.697532+00:00"},{"alias_kind":"pith_short_8","alias_value":"GV6322X3","created_at":"2026-07-05T08:17:06.697532+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2512.11321","citing_title":"KeyframeFace: Language-Driven Facial Animation via Semantic Keyframes","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2410.13720","citing_title":"Movie Gen: A Cast of Media Foundation Models","ref_index":35,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GV6322X3BC3SHZN5PWXZTTEPX3","json":"https://pith.science/pith/GV6322X3BC3SHZN5PWXZTTEPX3.json","graph_json":"https://pith.science/api/pith-number/GV6322X3BC3SHZN5PWXZTTEPX3/graph.json","events_json":"https://pith.science/api/pith-number/GV6322X3BC3SHZN5PWXZTTEPX3/events.json","paper":"https://pith.science/paper/GV6322X3"},"agent_actions":{"view_html":"https://pith.science/pith/GV6322X3BC3SHZN5PWXZTTEPX3","download_json":"https://pith.science/pith/GV6322X3BC3SHZN5PWXZTTEPX3.json","view_paper":"https://pith.science/paper/GV6322X3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.05224&json=true","fetch_graph":"https://pith.science/api/pith-number/GV6322X3BC3SHZN5PWXZTTEPX3/graph.json","fetch_events":"https://pith.science/api/pith-number/GV6322X3BC3SHZN5PWXZTTEPX3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GV6322X3BC3SHZN5PWXZTTEPX3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GV6322X3BC3SHZN5PWXZTTEPX3/action/storage_attestation","attest_author":"https://pith.science/pith/GV6322X3BC3SHZN5PWXZTTEPX3/action/author_attestation","sign_citation":"https://pith.science/pith/GV6322X3BC3SHZN5PWXZTTEPX3/action/citation_signature","submit_replication":"https://pith.science/pith/GV6322X3BC3SHZN5PWXZTTEPX3/action/replication_record"}},"created_at":"2026-07-05T08:17:06.697532+00:00","updated_at":"2026-07-05T08:17:06.697532+00:00"}