{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XKSBDS2Q2NENICCQTYNRR44MTA","short_pith_number":"pith:XKSBDS2Q","schema_version":"1.0","canonical_sha256":"baa411cb50d348d408509e1b18f38c98266abd7882b43d335cc0f561e327f3da","source":{"kind":"arxiv","id":"2502.01776","version":2},"attestation_state":"computed","paper":{"title":"Sparse VideoGen: Accelerating Video Diffusion Transformers with Spatial-Temporal Sparsity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Chenfeng Xu, Dacheng Li, Han Cai, Haocheng Xi, Ion Stoica, Jianfei Chen, Jintao Zhang, Kurt Keutzer, Muyang Li, Shuo Yang, Song Han, Xiuyu Li, Yilong Zhao, Yujun Lin","submitted_at":"2025-02-03T19:29:16Z","abstract_excerpt":"Diffusion Transformers (DiTs) dominate video generation but their high computational cost severely limits real-world applicability, usually requiring tens of minutes to generate a few seconds of video even on high-performance GPUs. This inefficiency primarily arises from the quadratic computational complexity of 3D Full Attention with respect to the context length. In this paper, we propose a training-free framework termed Sparse VideoGen (SVG) that leverages the inherent sparsity in 3D Full Attention to boost inference efficiency. We reveal that the attention heads can be dynamically classifi"},"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":"2502.01776","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-02-03T19:29:16Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"1a7fb0ba3bd9e826058c8279ab6a528606ab53f95042dad7a40ab092a9f302e7","abstract_canon_sha256":"8e64a1460cfc54b551043a619040800247176d0c2dc0e11122a79d266b053ed6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:54:42.702093Z","signature_b64":"clJRPYcVMBpTMTh58FxIZCj7yPq4iUnXmsVERYIOZNfDDF7UVQf4opcTGO08ljsYU4x+4T4Pz4MsZ+0waNxaBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"baa411cb50d348d408509e1b18f38c98266abd7882b43d335cc0f561e327f3da","last_reissued_at":"2026-07-05T10:54:42.701573Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:54:42.701573Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sparse VideoGen: Accelerating Video Diffusion Transformers with Spatial-Temporal Sparsity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Chenfeng Xu, Dacheng Li, Han Cai, Haocheng Xi, Ion Stoica, Jianfei Chen, Jintao Zhang, Kurt Keutzer, Muyang Li, Shuo Yang, Song Han, Xiuyu Li, Yilong Zhao, Yujun Lin","submitted_at":"2025-02-03T19:29:16Z","abstract_excerpt":"Diffusion Transformers (DiTs) dominate video generation but their high computational cost severely limits real-world applicability, usually requiring tens of minutes to generate a few seconds of video even on high-performance GPUs. This inefficiency primarily arises from the quadratic computational complexity of 3D Full Attention with respect to the context length. In this paper, we propose a training-free framework termed Sparse VideoGen (SVG) that leverages the inherent sparsity in 3D Full Attention to boost inference efficiency. We reveal that the attention heads can be dynamically classifi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.01776","kind":"arxiv","version":2},"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/2502.01776/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":"2502.01776","created_at":"2026-07-05T10:54:42.701635+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.01776v2","created_at":"2026-07-05T10:54:42.701635+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.01776","created_at":"2026-07-05T10:54:42.701635+00:00"},{"alias_kind":"pith_short_12","alias_value":"XKSBDS2Q2NEN","created_at":"2026-07-05T10:54:42.701635+00:00"},{"alias_kind":"pith_short_16","alias_value":"XKSBDS2Q2NENICCQ","created_at":"2026-07-05T10:54:42.701635+00:00"},{"alias_kind":"pith_short_8","alias_value":"XKSBDS2Q","created_at":"2026-07-05T10:54:42.701635+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":28,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.23019","citing_title":"ScalingAttention: Discovering Intrinsic Sparse Attention Topology for Video Diffusion Transformers","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23743","citing_title":"Sol Video Inference Engine: Agent-Native Full-Stack Acceleration Framework for Efficient Video Generation","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06309","citing_title":"RhymeFlow: Training-Free Acceleration for Video Generation with Asynchronous Denoising Flow Scheduling","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.31158","citing_title":"Light Interaction: Training-Free Inference Acceleration for Interactive Video World Models","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30849","citing_title":"SyncCache: Exploiting Asymmetric Dynamics for Fast Audio-Driven Portrait Animation","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2606.30557","citing_title":"EcoVideo: Entropy-Orchestrated Video Generation Paradigm in Cloud-Edge Dynamics","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26632","citing_title":"RT-Lynx: Putting the GEMM Sparsity In a Right Way for Diffusion Models","ref_index":65,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25798","citing_title":"DiSC: Resolution-Scalable Acceleration of Diffusion Models by Exploiting Sparsity and Cached Token Reuse with Hash-based Distribution","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28691","citing_title":"OSP-Next: Efficient High-Quality Video Generation with Sparse Sequence Parallelism, HiF8 Quantization, and Reinforcement Learning","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30325","citing_title":"Veda: Scalable Video Diffusion via Distilled Sparse Attention","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.31057","citing_title":"LVSA: Training-Free Sparse Attention for Long Video Diffusion","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23445","citing_title":"DFSAttn: Dynamic Fine-grained Sparse Attention for Efficient Video Generation","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2602.02214","citing_title":"Causal Forcing: Autoregressive Diffusion Distillation Done Right for High-Quality Real-Time Interactive Video Generation","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2603.21002","citing_title":"SURF: Signature-Retained Fast Video Generation","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2602.02214","citing_title":"Causal Forcing: Autoregressive Diffusion Distillation Done Right for High-Quality Real-Time Interactive Video Generation","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17633","citing_title":"SparseSAM: Structured Sparsification of Activations in Segment Anything Models","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2509.23980","citing_title":"Towards Redundancy Reduction in Diffusion Models for Efficient Video Super-Resolution","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2602.02958","citing_title":"Quant VideoGen: Auto-Regressive Long Video Generation via 2-Bit KV-Cache Quantization","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2602.22575","citing_title":"S2O: Early Stopping for Sparse Attention via Online Permutation","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2603.05811","citing_title":"Video Compression Meets Video Generation: Latent Inter-Frame Pruning with Attention Recovery","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2603.18636","citing_title":"Attention Sparsity is Input-Stable: Training-Free Sparse Attention for Video Generation via Offline Sparsity Profiling and Online QK Co-Clustering","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14513","citing_title":"HASTE: Training-Free Video Diffusion Acceleration via Head-Wise Adaptive Sparse Attention","ref_index":35,"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":32,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20470","citing_title":"DynamicRad: Content-Adaptive Sparse Attention for Long Video Diffusion","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10103","citing_title":"Long-Horizon Streaming Video Generation via Hybrid Attention with Decoupled Distillation","ref_index":47,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XKSBDS2Q2NENICCQTYNRR44MTA","json":"https://pith.science/pith/XKSBDS2Q2NENICCQTYNRR44MTA.json","graph_json":"https://pith.science/api/pith-number/XKSBDS2Q2NENICCQTYNRR44MTA/graph.json","events_json":"https://pith.science/api/pith-number/XKSBDS2Q2NENICCQTYNRR44MTA/events.json","paper":"https://pith.science/paper/XKSBDS2Q"},"agent_actions":{"view_html":"https://pith.science/pith/XKSBDS2Q2NENICCQTYNRR44MTA","download_json":"https://pith.science/pith/XKSBDS2Q2NENICCQTYNRR44MTA.json","view_paper":"https://pith.science/paper/XKSBDS2Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.01776&json=true","fetch_graph":"https://pith.science/api/pith-number/XKSBDS2Q2NENICCQTYNRR44MTA/graph.json","fetch_events":"https://pith.science/api/pith-number/XKSBDS2Q2NENICCQTYNRR44MTA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XKSBDS2Q2NENICCQTYNRR44MTA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XKSBDS2Q2NENICCQTYNRR44MTA/action/storage_attestation","attest_author":"https://pith.science/pith/XKSBDS2Q2NENICCQTYNRR44MTA/action/author_attestation","sign_citation":"https://pith.science/pith/XKSBDS2Q2NENICCQTYNRR44MTA/action/citation_signature","submit_replication":"https://pith.science/pith/XKSBDS2Q2NENICCQTYNRR44MTA/action/replication_record"}},"created_at":"2026-07-05T10:54:42.701635+00:00","updated_at":"2026-07-05T10:54:42.701635+00:00"}