{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:CXKZLBHJ6EDNVSC5P5CXLK2XDT","short_pith_number":"pith:CXKZLBHJ","schema_version":"1.0","canonical_sha256":"15d59584e9f106dac85d7f4575ab571cd04c8ca1cd2b6e788060eec290499c79","source":{"kind":"arxiv","id":"2402.13573","version":3},"attestation_state":"computed","paper":{"title":"ToDo: Token Downsampling for Efficient Generation of High-Resolution Images","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Aninda Saha, Ethan Smith, Nayan Saxena","submitted_at":"2024-02-21T07:10:28Z","abstract_excerpt":"Attention mechanism has been crucial for image diffusion models, however, their quadratic computational complexity limits the sizes of images we can process within reasonable time and memory constraints. This paper investigates the importance of dense attention in generative image models, which often contain redundant features, making them suitable for sparser attention mechanisms. We propose a novel training-free method ToDo that relies on token downsampling of key and value tokens to accelerate Stable Diffusion inference by up to 2x for common sizes and up to 4.5x or more for high resolution"},"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":"2402.13573","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-02-21T07:10:28Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"6feb3cd99be431a041e96804c62b34f694f4cd70a834548e0e1eb5723632789e","abstract_canon_sha256":"e38d3a9d7633bbfc857a05d3f0a9dc5b75587550381bdeb3cd2ce0cd75b472e7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:16:47.429715Z","signature_b64":"FaUouOxtQSkwPy9iDinna/vtqknL1vRMsrY6cRUBwI+07HLQEL0mDauOL4s+cxwBa9TarqfZ8fgFKbLANvmJAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"15d59584e9f106dac85d7f4575ab571cd04c8ca1cd2b6e788060eec290499c79","last_reissued_at":"2026-07-05T08:16:47.429209Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:16:47.429209Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ToDo: Token Downsampling for Efficient Generation of High-Resolution Images","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Aninda Saha, Ethan Smith, Nayan Saxena","submitted_at":"2024-02-21T07:10:28Z","abstract_excerpt":"Attention mechanism has been crucial for image diffusion models, however, their quadratic computational complexity limits the sizes of images we can process within reasonable time and memory constraints. This paper investigates the importance of dense attention in generative image models, which often contain redundant features, making them suitable for sparser attention mechanisms. We propose a novel training-free method ToDo that relies on token downsampling of key and value tokens to accelerate Stable Diffusion inference by up to 2x for common sizes and up to 4.5x or more for high resolution"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.13573","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/2402.13573/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":"2402.13573","created_at":"2026-07-05T08:16:47.429274+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.13573v3","created_at":"2026-07-05T08:16:47.429274+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.13573","created_at":"2026-07-05T08:16:47.429274+00:00"},{"alias_kind":"pith_short_12","alias_value":"CXKZLBHJ6EDN","created_at":"2026-07-05T08:16:47.429274+00:00"},{"alias_kind":"pith_short_16","alias_value":"CXKZLBHJ6EDNVSC5","created_at":"2026-07-05T08:16:47.429274+00:00"},{"alias_kind":"pith_short_8","alias_value":"CXKZLBHJ","created_at":"2026-07-05T08:16:47.429274+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.22011","citing_title":"Rethinking Token Reduction for Diffusion Models via Output-Similarity-Awareness","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14191","citing_title":"CoReDiT: Spatial Coherence-Guided Token Pruning and Reconstruction for Efficient Diffusion Transformers","ref_index":28,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CXKZLBHJ6EDNVSC5P5CXLK2XDT","json":"https://pith.science/pith/CXKZLBHJ6EDNVSC5P5CXLK2XDT.json","graph_json":"https://pith.science/api/pith-number/CXKZLBHJ6EDNVSC5P5CXLK2XDT/graph.json","events_json":"https://pith.science/api/pith-number/CXKZLBHJ6EDNVSC5P5CXLK2XDT/events.json","paper":"https://pith.science/paper/CXKZLBHJ"},"agent_actions":{"view_html":"https://pith.science/pith/CXKZLBHJ6EDNVSC5P5CXLK2XDT","download_json":"https://pith.science/pith/CXKZLBHJ6EDNVSC5P5CXLK2XDT.json","view_paper":"https://pith.science/paper/CXKZLBHJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.13573&json=true","fetch_graph":"https://pith.science/api/pith-number/CXKZLBHJ6EDNVSC5P5CXLK2XDT/graph.json","fetch_events":"https://pith.science/api/pith-number/CXKZLBHJ6EDNVSC5P5CXLK2XDT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CXKZLBHJ6EDNVSC5P5CXLK2XDT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CXKZLBHJ6EDNVSC5P5CXLK2XDT/action/storage_attestation","attest_author":"https://pith.science/pith/CXKZLBHJ6EDNVSC5P5CXLK2XDT/action/author_attestation","sign_citation":"https://pith.science/pith/CXKZLBHJ6EDNVSC5P5CXLK2XDT/action/citation_signature","submit_replication":"https://pith.science/pith/CXKZLBHJ6EDNVSC5P5CXLK2XDT/action/replication_record"}},"created_at":"2026-07-05T08:16:47.429274+00:00","updated_at":"2026-07-05T08:16:47.429274+00:00"}