{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:QDWH4FYSKHFSJLVREWN4PU27N2","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"f04225a61e43de816dd86645f93e02d3bb37c9e5758932130e81af872c72a596","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-14T07:57:55Z","title_canon_sha256":"b9227144d45ce8fd31aefe215632180d3954b4c277575990c45d9044dae0a580"},"schema_version":"1.0","source":{"id":"2605.14513","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14513","created_at":"2026-05-17T23:39:06Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14513v1","created_at":"2026-05-17T23:39:06Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14513","created_at":"2026-05-17T23:39:06Z"},{"alias_kind":"pith_short_12","alias_value":"QDWH4FYSKHFS","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"QDWH4FYSKHFSJLVR","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"QDWH4FYS","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:a156a2557654007fef07aa3cb0a31077f709f9a6fb04ef7d6622d5a38eac1ae8","target":"graph","created_at":"2026-05-17T23:39:06Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"On Wan2.1-1.3B and Wan2.1-14B, our method consistently improves XAttention and SVG2, achieving up to 1.93 times speedup at 720P while maintaining competitive video quality and similarity metrics."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That measured model-output error under a global sparsity budget reliably predicts perceptual video quality across heads and that temporal query-key drift is stable enough for safe mask reuse without visible artifacts."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"HASTE delivers up to 1.93x speedup on Wan2.1 video DiTs via head-wise adaptive sparse attention using temporal mask reuse and error-guided per-head calibration while preserving video quality."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Head-wise adaptive sparse attention accelerates pretrained video diffusion models up to 1.93 times without retraining by reusing temporal masks and calibrating sparsity per head."}],"snapshot_sha256":"12881d9dd2943bf8b402cebdcd01c9ee463c129c904b2ad940caa4efa71b338f"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Diffusion-based video generation has advanced substantially in visual fidelity and temporal coherence, but practical deployment remains limited by the quadratic complexity of full attention. Training-free sparse attention is attractive because it accelerates pretrained models without retraining, yet existing online top-$p$ sparse attention still spends non-negligible cost on mask prediction and applies shared thresholds despite strong head-level heterogeneity. We show that these two overlooked factors limit the practical speed-quality trade-off of training-free sparse attention in Video DiTs. ","authors_text":"Fei Chao, Jing Xu, Rongrong Ji, Xiawu Zheng, Xuzhe Zheng, Yuexiao Ma","cross_cats":["cs.AI"],"headline":"Head-wise adaptive sparse attention accelerates pretrained video diffusion models up to 1.93 times without retraining by reusing temporal masks and calibrating sparsity per head.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-14T07:57:55Z","title":"HASTE: Training-Free Video Diffusion Acceleration via Head-Wise Adaptive Sparse Attention"},"references":{"count":55,"internal_anchors":11,"resolved_work":55,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Dicache: Let diffusion model determine its own cache","work_id":"c2389322-8839-490a-8d61-34832e66ebb6","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Rainfusion: Adaptive video generation acceleration via multi-dimensional visual redundancy","work_id":"a6bfc322-49cb-4f3c-bf04-19fcfb1e39b5","year":2025},{"cited_arxiv_id":"2512.24086","doi":"","is_internal_anchor":true,"ref_index":3,"title":"RainFusion2.0: Temporal-Spatial Awareness and Hardware-Efficient Block-wise Sparse Attention","work_id":"179d3710-8907-440f-8695-8d1a532bfe14","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Sparse-vDiT: Unleashing the Power of Sparse Attention to Accelerate Video Diffusion Transformers, June 2025","work_id":"8ff929a3-a68f-4193-909e-1b19fbdb4f57","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Hicache: Training-free acceleration of diffusion models via hermite polynomial- based feature caching.arXiv preprint arXiv:2508.16984","work_id":"43ebda46-bf78-416c-9a3e-a372d3d1b9e7","year":2025}],"snapshot_sha256":"de4fd7557ab4c60644eb972a4e21a505b7104439c8879b76ffea616e27d90624"},"source":{"id":"2605.14513","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-15T05:14:18.704869Z","id":"743a148e-dd17-4d7b-abeb-b7248690224f","model_set":{"reader":"grok-4.3"},"one_line_summary":"HASTE delivers up to 1.93x speedup on Wan2.1 video DiTs via head-wise adaptive sparse attention using temporal mask reuse and error-guided per-head calibration while preserving video quality.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Head-wise adaptive sparse attention accelerates pretrained video diffusion models up to 1.93 times without retraining by reusing temporal masks and calibrating sparsity per head.","strongest_claim":"On Wan2.1-1.3B and Wan2.1-14B, our method consistently improves XAttention and SVG2, achieving up to 1.93 times speedup at 720P while maintaining competitive video quality and similarity metrics.","weakest_assumption":"That measured model-output error under a global sparsity budget reliably predicts perceptual video quality across heads and that temporal query-key drift is stable enough for safe mask reuse without visible artifacts."}},"verdict_id":"743a148e-dd17-4d7b-abeb-b7248690224f"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:948a91f13dde67f46ac6a8279b92d4cc02c53fea98d9dc111ea01c3317745be8","target":"record","created_at":"2026-05-17T23:39:06Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"f04225a61e43de816dd86645f93e02d3bb37c9e5758932130e81af872c72a596","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-14T07:57:55Z","title_canon_sha256":"b9227144d45ce8fd31aefe215632180d3954b4c277575990c45d9044dae0a580"},"schema_version":"1.0","source":{"id":"2605.14513","kind":"arxiv","version":1}},"canonical_sha256":"80ec7e171251cb24aeb1259bc7d35f6e837a4669680ba7991c4b865869cf8992","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"80ec7e171251cb24aeb1259bc7d35f6e837a4669680ba7991c4b865869cf8992","first_computed_at":"2026-05-17T23:39:06.168512Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:06.168512Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"G5eMY8QMxLsIjJevqVRxDhMvh43eeG8+G1QrIJR1lLlnwgqzdZ1B6eMW0cHk6Zq/gqMdsjGJolmnAX/FPx22Bg==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:06.169103Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.14513","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:948a91f13dde67f46ac6a8279b92d4cc02c53fea98d9dc111ea01c3317745be8","sha256:a156a2557654007fef07aa3cb0a31077f709f9a6fb04ef7d6622d5a38eac1ae8"],"state_sha256":"152220e87e19a6ee8cdf27f2978c0f6eed29eb0c48db59a4802c4f7621a360e8"}