{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:NB4ESE4V6NWW3CP6QX5KJA5UQ3","short_pith_number":"pith:NB4ESE4V","schema_version":"1.0","canonical_sha256":"6878491395f36d6d89fe85faa483b486e8687f25ba821235e63420f0c58e6c9a","source":{"kind":"arxiv","id":"2412.16112","version":1},"attestation_state":"computed","paper":{"title":"CLEAR: Conv-Like Linearization Revs Pre-Trained Diffusion Transformers Up","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Songhua Liu, Xinchao Wang, Zhenxiong Tan","submitted_at":"2024-12-20T17:57:09Z","abstract_excerpt":"Diffusion Transformers (DiT) have become a leading architecture in image generation. However, the quadratic complexity of attention mechanisms, which are responsible for modeling token-wise relationships, results in significant latency when generating high-resolution images. To address this issue, we aim at a linear attention mechanism in this paper that reduces the complexity of pre-trained DiTs to linear. We begin our exploration with a comprehensive summary of existing efficient attention mechanisms and identify four key factors crucial for successful linearization of pre-trained DiTs: loca"},"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":"2412.16112","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-12-20T17:57:09Z","cross_cats_sorted":[],"title_canon_sha256":"ca0aace8be4e98d655e941171c09a5a39942a66ba78ff6a7b3c9b531557eccb8","abstract_canon_sha256":"83edb8ffc6a4eb079734e9adc99ac99c6a9d5dd631a978e47c3089f421bc0e43"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:52:35.214590Z","signature_b64":"Jt4WqFJ3fK0WxIeshGDw8iSXoR6McApMGOoA+23Yy9iCMBXn9uBh7oLA4doHok4FsYv+hdvioafRjxMVXC9sAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6878491395f36d6d89fe85faa483b486e8687f25ba821235e63420f0c58e6c9a","last_reissued_at":"2026-07-05T09:52:35.214157Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:52:35.214157Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CLEAR: Conv-Like Linearization Revs Pre-Trained Diffusion Transformers Up","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Songhua Liu, Xinchao Wang, Zhenxiong Tan","submitted_at":"2024-12-20T17:57:09Z","abstract_excerpt":"Diffusion Transformers (DiT) have become a leading architecture in image generation. However, the quadratic complexity of attention mechanisms, which are responsible for modeling token-wise relationships, results in significant latency when generating high-resolution images. To address this issue, we aim at a linear attention mechanism in this paper that reduces the complexity of pre-trained DiTs to linear. We begin our exploration with a comprehensive summary of existing efficient attention mechanisms and identify four key factors crucial for successful linearization of pre-trained DiTs: loca"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.16112","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/2412.16112/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":"2412.16112","created_at":"2026-07-05T09:52:35.214215+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.16112v1","created_at":"2026-07-05T09:52:35.214215+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.16112","created_at":"2026-07-05T09:52:35.214215+00:00"},{"alias_kind":"pith_short_12","alias_value":"NB4ESE4V6NWW","created_at":"2026-07-05T09:52:35.214215+00:00"},{"alias_kind":"pith_short_16","alias_value":"NB4ESE4V6NWW3CP6","created_at":"2026-07-05T09:52:35.214215+00:00"},{"alias_kind":"pith_short_8","alias_value":"NB4ESE4V","created_at":"2026-07-05T09:52:35.214215+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.14191","citing_title":"CoReDiT: Spatial Coherence-Guided Token Pruning and Reconstruction for Efficient Diffusion Transformers","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02772","citing_title":"Linearizing Vision Transformer with Test-Time Training","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NB4ESE4V6NWW3CP6QX5KJA5UQ3","json":"https://pith.science/pith/NB4ESE4V6NWW3CP6QX5KJA5UQ3.json","graph_json":"https://pith.science/api/pith-number/NB4ESE4V6NWW3CP6QX5KJA5UQ3/graph.json","events_json":"https://pith.science/api/pith-number/NB4ESE4V6NWW3CP6QX5KJA5UQ3/events.json","paper":"https://pith.science/paper/NB4ESE4V"},"agent_actions":{"view_html":"https://pith.science/pith/NB4ESE4V6NWW3CP6QX5KJA5UQ3","download_json":"https://pith.science/pith/NB4ESE4V6NWW3CP6QX5KJA5UQ3.json","view_paper":"https://pith.science/paper/NB4ESE4V","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.16112&json=true","fetch_graph":"https://pith.science/api/pith-number/NB4ESE4V6NWW3CP6QX5KJA5UQ3/graph.json","fetch_events":"https://pith.science/api/pith-number/NB4ESE4V6NWW3CP6QX5KJA5UQ3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NB4ESE4V6NWW3CP6QX5KJA5UQ3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NB4ESE4V6NWW3CP6QX5KJA5UQ3/action/storage_attestation","attest_author":"https://pith.science/pith/NB4ESE4V6NWW3CP6QX5KJA5UQ3/action/author_attestation","sign_citation":"https://pith.science/pith/NB4ESE4V6NWW3CP6QX5KJA5UQ3/action/citation_signature","submit_replication":"https://pith.science/pith/NB4ESE4V6NWW3CP6QX5KJA5UQ3/action/replication_record"}},"created_at":"2026-07-05T09:52:35.214215+00:00","updated_at":"2026-07-05T09:52:35.214215+00:00"}