{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:WSUWUA7PJTGHMAXT2GHXFSPRTH","short_pith_number":"pith:WSUWUA7P","schema_version":"1.0","canonical_sha256":"b4a96a03ef4ccc7602f3d18f72c9f199c995b6746ebce02d64b45e833f8e44e0","source":{"kind":"arxiv","id":"2405.17403","version":3},"attestation_state":"computed","paper":{"title":"A Closer Look at Time Steps is Worthy of Triple Speed-Up for Diffusion Model Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hanwang Zhang, Kai Wang, Mingjia Shi, Xiaojiang Peng, Yang You, Yukun Zhou, Yuzhang Shang, Zekai Li, Zhihang Yuan","submitted_at":"2024-05-27T17:51:36Z","abstract_excerpt":"Training diffusion models is always a computation-intensive task. In this paper, we introduce a novel speed-up method for diffusion model training, called, which is based on a closer look at time steps. Our key findings are: i) Time steps can be empirically divided into acceleration, deceleration, and convergence areas based on the process increment. ii) These time steps are imbalanced, with many concentrated in the convergence area. iii) The concentrated steps provide limited benefits for diffusion training. To address this, we design an asymmetric sampling strategy that reduces the frequency"},"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":"2405.17403","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-05-27T17:51:36Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"0a8b33402fb0509b31b49446acd0adcd62b5480c7b3933aa09a6b78384213457","abstract_canon_sha256":"01458bc3dd5ddd7e25a5ee76628fc82bd7059b1cdad762abf2aeb6212130d450"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:38:33.562062Z","signature_b64":"lHYqB6RTsYK/hv9NA/lxRj4lUZGLT5YxnNP+9wOWuTSpueAtTPJjhtG/KbOxsnld0mhRDnwne56tB7/x2r5eDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b4a96a03ef4ccc7602f3d18f72c9f199c995b6746ebce02d64b45e833f8e44e0","last_reissued_at":"2026-07-05T10:38:33.561573Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:38:33.561573Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Closer Look at Time Steps is Worthy of Triple Speed-Up for Diffusion Model Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hanwang Zhang, Kai Wang, Mingjia Shi, Xiaojiang Peng, Yang You, Yukun Zhou, Yuzhang Shang, Zekai Li, Zhihang Yuan","submitted_at":"2024-05-27T17:51:36Z","abstract_excerpt":"Training diffusion models is always a computation-intensive task. In this paper, we introduce a novel speed-up method for diffusion model training, called, which is based on a closer look at time steps. Our key findings are: i) Time steps can be empirically divided into acceleration, deceleration, and convergence areas based on the process increment. ii) These time steps are imbalanced, with many concentrated in the convergence area. iii) The concentrated steps provide limited benefits for diffusion training. To address this, we design an asymmetric sampling strategy that reduces the frequency"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.17403","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/2405.17403/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":"2405.17403","created_at":"2026-07-05T10:38:33.561632+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.17403v3","created_at":"2026-07-05T10:38:33.561632+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.17403","created_at":"2026-07-05T10:38:33.561632+00:00"},{"alias_kind":"pith_short_12","alias_value":"WSUWUA7PJTGH","created_at":"2026-07-05T10:38:33.561632+00:00"},{"alias_kind":"pith_short_16","alias_value":"WSUWUA7PJTGHMAXT","created_at":"2026-07-05T10:38:33.561632+00:00"},{"alias_kind":"pith_short_8","alias_value":"WSUWUA7P","created_at":"2026-07-05T10:38:33.561632+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.10790","citing_title":"Elucidating Representation Degradation Problem in Diffusion Model Training","ref_index":55,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WSUWUA7PJTGHMAXT2GHXFSPRTH","json":"https://pith.science/pith/WSUWUA7PJTGHMAXT2GHXFSPRTH.json","graph_json":"https://pith.science/api/pith-number/WSUWUA7PJTGHMAXT2GHXFSPRTH/graph.json","events_json":"https://pith.science/api/pith-number/WSUWUA7PJTGHMAXT2GHXFSPRTH/events.json","paper":"https://pith.science/paper/WSUWUA7P"},"agent_actions":{"view_html":"https://pith.science/pith/WSUWUA7PJTGHMAXT2GHXFSPRTH","download_json":"https://pith.science/pith/WSUWUA7PJTGHMAXT2GHXFSPRTH.json","view_paper":"https://pith.science/paper/WSUWUA7P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.17403&json=true","fetch_graph":"https://pith.science/api/pith-number/WSUWUA7PJTGHMAXT2GHXFSPRTH/graph.json","fetch_events":"https://pith.science/api/pith-number/WSUWUA7PJTGHMAXT2GHXFSPRTH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WSUWUA7PJTGHMAXT2GHXFSPRTH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WSUWUA7PJTGHMAXT2GHXFSPRTH/action/storage_attestation","attest_author":"https://pith.science/pith/WSUWUA7PJTGHMAXT2GHXFSPRTH/action/author_attestation","sign_citation":"https://pith.science/pith/WSUWUA7PJTGHMAXT2GHXFSPRTH/action/citation_signature","submit_replication":"https://pith.science/pith/WSUWUA7PJTGHMAXT2GHXFSPRTH/action/replication_record"}},"created_at":"2026-07-05T10:38:33.561632+00:00","updated_at":"2026-07-05T10:38:33.561632+00:00"}