{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:4HKYOHBSR36CD764OHNZDAEEFN","short_pith_number":"pith:4HKYOHBS","schema_version":"1.0","canonical_sha256":"e1d5871c328efc21ffdc71db9180842b5f65207ecbb25eb55f7f38ccdcabef5b","source":{"kind":"arxiv","id":"2410.13276","version":4},"attestation_state":"computed","paper":{"title":"SeerAttention: Learning Intrinsic Sparse Attention in Your LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dayou Du, Fan Yang, Hayden Kwok-Hay So, Jiaxing Qi, Junjie Lai, Mao Yang, Peiyuan Zhou, Shijie Cao, Ting Cao, Yizhao Gao, Zhichen Zeng","submitted_at":"2024-10-17T07:07:09Z","abstract_excerpt":"Attention is the cornerstone of modern Large Language Models (LLMs). Yet its quadratic complexity hinders efficiency and scalability, especially for long-context processing. A promising approach is to leverage sparsity in attention. However, existing sparsity-based solutions predominantly rely on predefined patterns or heuristics at the attention head level, struggling to adapt dynamically to different contexts efficiently.\n  We propose SeerAttention, a simple yet effective attention mechanism that directly learns the block-level attention sparsity from the LLM itself. Inspired by the gating m"},"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":"2410.13276","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-17T07:07:09Z","cross_cats_sorted":[],"title_canon_sha256":"0d7a11eefae4a9fb00d4c118ac9bb61dd4fa30c028229fb675f20fe21647bdcd","abstract_canon_sha256":"4fac78567c558bcc0c4e64684980f6601b339254f1475101fcad4126244d5e3d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:15:20.675478Z","signature_b64":"SOJoyS0p9rWmTt7S6Osh6Ikt7k1Sn/hp8klTppZxAhVCmqB4Y7+iaYK70W22ONKMq6h/nFJDP8rIod+cv7T4DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e1d5871c328efc21ffdc71db9180842b5f65207ecbb25eb55f7f38ccdcabef5b","last_reissued_at":"2026-07-05T10:15:20.674946Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:15:20.674946Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SeerAttention: Learning Intrinsic Sparse Attention in Your LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dayou Du, Fan Yang, Hayden Kwok-Hay So, Jiaxing Qi, Junjie Lai, Mao Yang, Peiyuan Zhou, Shijie Cao, Ting Cao, Yizhao Gao, Zhichen Zeng","submitted_at":"2024-10-17T07:07:09Z","abstract_excerpt":"Attention is the cornerstone of modern Large Language Models (LLMs). Yet its quadratic complexity hinders efficiency and scalability, especially for long-context processing. A promising approach is to leverage sparsity in attention. However, existing sparsity-based solutions predominantly rely on predefined patterns or heuristics at the attention head level, struggling to adapt dynamically to different contexts efficiently.\n  We propose SeerAttention, a simple yet effective attention mechanism that directly learns the block-level attention sparsity from the LLM itself. Inspired by the gating m"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.13276","kind":"arxiv","version":4},"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/2410.13276/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":"2410.13276","created_at":"2026-07-05T10:15:20.675004+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.13276v4","created_at":"2026-07-05T10:15:20.675004+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.13276","created_at":"2026-07-05T10:15:20.675004+00:00"},{"alias_kind":"pith_short_12","alias_value":"4HKYOHBSR36C","created_at":"2026-07-05T10:15:20.675004+00:00"},{"alias_kind":"pith_short_16","alias_value":"4HKYOHBSR36CD764","created_at":"2026-07-05T10:15:20.675004+00:00"},{"alias_kind":"pith_short_8","alias_value":"4HKYOHBS","created_at":"2026-07-05T10:15:20.675004+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":16,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08032","citing_title":"What to Keep, What to Forget: A Rate--Distortion View of Memory Compaction in LLMs and Agents","ref_index":40,"is_internal_anchor":true},{"citing_arxiv_id":"2606.06467","citing_title":"You Only Index Once: Cross-Layer Sparse Attention with Shared Routing","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30325","citing_title":"Veda: Scalable Video Diffusion via Distilled Sparse Attention","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2509.07120","citing_title":"Block-Sparse Global Attention for Efficient Multi-View Geometry Transformers","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18753","citing_title":"DashAttention: Differentiable and Adaptive Sparse Hierarchical Attention","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16839","citing_title":"CompactAttention: Accelerating Chunked Prefill with Block-Union KV Selection","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2502.11089","citing_title":"Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2502.13189","citing_title":"MoBA: Mixture of Block Attention for Long-Context LLMs","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13080","citing_title":"Learning to See What You Need: Gaze Attention for Multimodal Large Language Models","ref_index":183,"is_internal_anchor":false},{"citing_arxiv_id":"2603.29002","citing_title":"Understand and Accelerate Memory Processing Pipeline for Large Language Model Inference","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24820","citing_title":"Salca: A Sparsity-Aware Hardware Accelerator for Efficient Long-Context Attention Decoding","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07394","citing_title":"Flux Attention: Context-Aware Hybrid Attention for Efficient LLMs Inference","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07363","citing_title":"MISA: Mixture of Indexer Sparse Attention for Long-Context LLM Inference","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07719","citing_title":"An Efficient Hybrid Sparse Attention with CPU-GPU Parallelism for Long-Context Inference","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14925","citing_title":"Improving Sparse Autoencoder with Dynamic Attention","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2505.02922","citing_title":"RetroInfer: A Vector Storage Engine for Scalable Long-Context LLM Inference","ref_index":31,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4HKYOHBSR36CD764OHNZDAEEFN","json":"https://pith.science/pith/4HKYOHBSR36CD764OHNZDAEEFN.json","graph_json":"https://pith.science/api/pith-number/4HKYOHBSR36CD764OHNZDAEEFN/graph.json","events_json":"https://pith.science/api/pith-number/4HKYOHBSR36CD764OHNZDAEEFN/events.json","paper":"https://pith.science/paper/4HKYOHBS"},"agent_actions":{"view_html":"https://pith.science/pith/4HKYOHBSR36CD764OHNZDAEEFN","download_json":"https://pith.science/pith/4HKYOHBSR36CD764OHNZDAEEFN.json","view_paper":"https://pith.science/paper/4HKYOHBS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.13276&json=true","fetch_graph":"https://pith.science/api/pith-number/4HKYOHBSR36CD764OHNZDAEEFN/graph.json","fetch_events":"https://pith.science/api/pith-number/4HKYOHBSR36CD764OHNZDAEEFN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4HKYOHBSR36CD764OHNZDAEEFN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4HKYOHBSR36CD764OHNZDAEEFN/action/storage_attestation","attest_author":"https://pith.science/pith/4HKYOHBSR36CD764OHNZDAEEFN/action/author_attestation","sign_citation":"https://pith.science/pith/4HKYOHBSR36CD764OHNZDAEEFN/action/citation_signature","submit_replication":"https://pith.science/pith/4HKYOHBSR36CD764OHNZDAEEFN/action/replication_record"}},"created_at":"2026-07-05T10:15:20.675004+00:00","updated_at":"2026-07-05T10:15:20.675004+00:00"}