{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:YNOANHRBSCQIMPUXYNAZQIG6Z6","short_pith_number":"pith:YNOANHRB","schema_version":"1.0","canonical_sha256":"c35c069e2190a0863e97c3419820decfa4b35f7aaf9f85844bd858094aeb8dbf","source":{"kind":"arxiv","id":"2404.02747","version":3},"attestation_state":"computed","paper":{"title":"Faster Diffusion via Temporal Attention Decomposition","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Francesco Faccio, Haozhe Liu, Jinheng Xie, Juan-Manuel Perez-Rua, J\\\"urgen Schmidhuber, Mengmeng Xu, Mike Zheng Shou, Tao Xiang, Wentian Zhang","submitted_at":"2024-04-03T13:44:41Z","abstract_excerpt":"We explore the role of attention mechanism during inference in text-conditional diffusion models. Empirical observations suggest that cross-attention outputs converge to a fixed point after several inference steps. The convergence time naturally divides the entire inference process into two phases: an initial phase for planning text-oriented visual semantics, which are then translated into images in a subsequent fidelity-improving phase. Cross-attention is essential in the initial phase but almost irrelevant thereafter. However, self-attention initially plays a minor role but becomes crucial i"},"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":"2404.02747","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-04-03T13:44:41Z","cross_cats_sorted":[],"title_canon_sha256":"879a1b5b7301a20e35371914da3080ef042172e2c682ef01416c2053780660a7","abstract_canon_sha256":"cd67c2e79c6e1d2be4b04519c1e75a1067a9c8170f3c17cc75b98b87ba1b018d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:20:11.247412Z","signature_b64":"UlbvDNsID6auPInRVHQq2CysAeefPcmRrqURdEbrsdJ4dx3QX5fQaK9ZCiAOyx5LEok9dj/Tz6wubLR8BePRAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c35c069e2190a0863e97c3419820decfa4b35f7aaf9f85844bd858094aeb8dbf","last_reissued_at":"2026-07-05T10:20:11.246904Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:20:11.246904Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Faster Diffusion via Temporal Attention Decomposition","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Francesco Faccio, Haozhe Liu, Jinheng Xie, Juan-Manuel Perez-Rua, J\\\"urgen Schmidhuber, Mengmeng Xu, Mike Zheng Shou, Tao Xiang, Wentian Zhang","submitted_at":"2024-04-03T13:44:41Z","abstract_excerpt":"We explore the role of attention mechanism during inference in text-conditional diffusion models. Empirical observations suggest that cross-attention outputs converge to a fixed point after several inference steps. The convergence time naturally divides the entire inference process into two phases: an initial phase for planning text-oriented visual semantics, which are then translated into images in a subsequent fidelity-improving phase. Cross-attention is essential in the initial phase but almost irrelevant thereafter. However, self-attention initially plays a minor role but becomes crucial i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.02747","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/2404.02747/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":"2404.02747","created_at":"2026-07-05T10:20:11.246959+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.02747v3","created_at":"2026-07-05T10:20:11.246959+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.02747","created_at":"2026-07-05T10:20:11.246959+00:00"},{"alias_kind":"pith_short_12","alias_value":"YNOANHRBSCQI","created_at":"2026-07-05T10:20:11.246959+00:00"},{"alias_kind":"pith_short_16","alias_value":"YNOANHRBSCQIMPUX","created_at":"2026-07-05T10:20:11.246959+00:00"},{"alias_kind":"pith_short_8","alias_value":"YNOANHRB","created_at":"2026-07-05T10:20:11.246959+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05319","citing_title":"Steering Optimisation Trajectories in Diffusion Representation Learning","ref_index":50,"is_internal_anchor":true},{"citing_arxiv_id":"2606.17378","citing_title":"RISE: Relay Inference and Online Scheduling for Efficient Edge-Device Collaborative Diffusion Model Services","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06060","citing_title":"ReCache: Learning Budget-Aware Caching Schedules for Diffusion Models via REINFORCE","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09313","citing_title":"Attention Sinks in Diffusion Transformers: A Causal Analysis","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31026","citing_title":"OTCache: Optimal Transport for Geometry-Aware Caching in Diffusion Models","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23381","citing_title":"VDE: Training-Free Accelerating Rectified Flow Model via Velocity Decomposition and Estimation","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22015","citing_title":"ORBIS: Output-Guided Token Reduction with Distribution-Aware Matching for Video Diffusion Acceleration","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18190","citing_title":"Dual-Rate Diffusion: Accelerating diffusion models with an interleaved heavy-light network","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06052","citing_title":"Attention, May I Have Your Decision? Localizing Generative Choices in Diffusion Models","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09313","citing_title":"Attention Sinks in Diffusion Transformers: A Causal Analysis","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09313","citing_title":"Attention Sinks in Diffusion Transformers: A Causal Analysis","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20470","citing_title":"DynamicRad: Content-Adaptive Sparse Attention for Long Video Diffusion","ref_index":59,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YNOANHRBSCQIMPUXYNAZQIG6Z6","json":"https://pith.science/pith/YNOANHRBSCQIMPUXYNAZQIG6Z6.json","graph_json":"https://pith.science/api/pith-number/YNOANHRBSCQIMPUXYNAZQIG6Z6/graph.json","events_json":"https://pith.science/api/pith-number/YNOANHRBSCQIMPUXYNAZQIG6Z6/events.json","paper":"https://pith.science/paper/YNOANHRB"},"agent_actions":{"view_html":"https://pith.science/pith/YNOANHRBSCQIMPUXYNAZQIG6Z6","download_json":"https://pith.science/pith/YNOANHRBSCQIMPUXYNAZQIG6Z6.json","view_paper":"https://pith.science/paper/YNOANHRB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.02747&json=true","fetch_graph":"https://pith.science/api/pith-number/YNOANHRBSCQIMPUXYNAZQIG6Z6/graph.json","fetch_events":"https://pith.science/api/pith-number/YNOANHRBSCQIMPUXYNAZQIG6Z6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YNOANHRBSCQIMPUXYNAZQIG6Z6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YNOANHRBSCQIMPUXYNAZQIG6Z6/action/storage_attestation","attest_author":"https://pith.science/pith/YNOANHRBSCQIMPUXYNAZQIG6Z6/action/author_attestation","sign_citation":"https://pith.science/pith/YNOANHRBSCQIMPUXYNAZQIG6Z6/action/citation_signature","submit_replication":"https://pith.science/pith/YNOANHRBSCQIMPUXYNAZQIG6Z6/action/replication_record"}},"created_at":"2026-07-05T10:20:11.246959+00:00","updated_at":"2026-07-05T10:20:11.246959+00:00"}