{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:42UHATZWMERTTQALRGWVLDKSM3","short_pith_number":"pith:42UHATZW","schema_version":"1.0","canonical_sha256":"e6a8704f36612339c00b89ad558d5266c63d9a06b3d176f2f0da96b089c8055f","source":{"kind":"arxiv","id":"2410.16135","version":3},"attestation_state":"computed","paper":{"title":"Beyond 2:4: exploring V:N:M sparsity for efficient transformer inference on GPUs","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chang Gao, Han Bao, Jianfei Chen, Kang Zhao, Liping Jing, Tao Yuan, Zhaofeng Sun, Zhenfeng Su, Zichen Liang","submitted_at":"2024-10-21T16:00:04Z","abstract_excerpt":"To date, 2:4 sparsity has stood as the only sparse pattern that can be accelerated using sparse tensor cores on GPUs. In practice, 2:4 sparsity often possesses low actual speedups ($\\leq 1.3$) and requires fixed sparse ratios, meaning that other ratios, such as 4:8, 8:16, or those exceeding 50% sparsity, do not incur any speedups on GPUs. Recent studies suggest that V:N:M sparsity is promising in addressing these limitations of 2:4 sparsity. However, regarding accuracy, the effects of V:N:M sparsity on broader Transformer models, such as vision Transformers and large language models (LLMs), ar"},"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":"2410.16135","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-21T16:00:04Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a214b7a9dcf9103b387e57efb7652590bd527e27c8de6143ccad6b95b978c582","abstract_canon_sha256":"47237c989fd1e16148be3d97c9c69f2ad74fdaaa015269dec3681b793ee6cba4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:14:25.372410Z","signature_b64":"o4kmsMw5yai1mMgZSjsdNNw90daJ1KO8knNQdMzDkV8PRfJPLqdC9qybJRJr6UOKvUgvVDRrlftymWCYYjRnDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e6a8704f36612339c00b89ad558d5266c63d9a06b3d176f2f0da96b089c8055f","last_reissued_at":"2026-07-05T11:14:25.371982Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:14:25.371982Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond 2:4: exploring V:N:M sparsity for efficient transformer inference on GPUs","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chang Gao, Han Bao, Jianfei Chen, Kang Zhao, Liping Jing, Tao Yuan, Zhaofeng Sun, Zhenfeng Su, Zichen Liang","submitted_at":"2024-10-21T16:00:04Z","abstract_excerpt":"To date, 2:4 sparsity has stood as the only sparse pattern that can be accelerated using sparse tensor cores on GPUs. In practice, 2:4 sparsity often possesses low actual speedups ($\\leq 1.3$) and requires fixed sparse ratios, meaning that other ratios, such as 4:8, 8:16, or those exceeding 50% sparsity, do not incur any speedups on GPUs. Recent studies suggest that V:N:M sparsity is promising in addressing these limitations of 2:4 sparsity. However, regarding accuracy, the effects of V:N:M sparsity on broader Transformer models, such as vision Transformers and large language models (LLMs), ar"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.16135","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/2410.16135/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":"2410.16135","created_at":"2026-07-05T11:14:25.372040+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.16135v3","created_at":"2026-07-05T11:14:25.372040+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.16135","created_at":"2026-07-05T11:14:25.372040+00:00"},{"alias_kind":"pith_short_12","alias_value":"42UHATZWMERT","created_at":"2026-07-05T11:14:25.372040+00:00"},{"alias_kind":"pith_short_16","alias_value":"42UHATZWMERTTQAL","created_at":"2026-07-05T11:14:25.372040+00:00"},{"alias_kind":"pith_short_8","alias_value":"42UHATZW","created_at":"2026-07-05T11:14:25.372040+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2507.03052","citing_title":"From 2:4 to 8:16 sparsity patterns in LLMs for Outliers and Weights with Variance Correction","ref_index":33,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/42UHATZWMERTTQALRGWVLDKSM3","json":"https://pith.science/pith/42UHATZWMERTTQALRGWVLDKSM3.json","graph_json":"https://pith.science/api/pith-number/42UHATZWMERTTQALRGWVLDKSM3/graph.json","events_json":"https://pith.science/api/pith-number/42UHATZWMERTTQALRGWVLDKSM3/events.json","paper":"https://pith.science/paper/42UHATZW"},"agent_actions":{"view_html":"https://pith.science/pith/42UHATZWMERTTQALRGWVLDKSM3","download_json":"https://pith.science/pith/42UHATZWMERTTQALRGWVLDKSM3.json","view_paper":"https://pith.science/paper/42UHATZW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.16135&json=true","fetch_graph":"https://pith.science/api/pith-number/42UHATZWMERTTQALRGWVLDKSM3/graph.json","fetch_events":"https://pith.science/api/pith-number/42UHATZWMERTTQALRGWVLDKSM3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/42UHATZWMERTTQALRGWVLDKSM3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/42UHATZWMERTTQALRGWVLDKSM3/action/storage_attestation","attest_author":"https://pith.science/pith/42UHATZWMERTTQALRGWVLDKSM3/action/author_attestation","sign_citation":"https://pith.science/pith/42UHATZWMERTTQALRGWVLDKSM3/action/citation_signature","submit_replication":"https://pith.science/pith/42UHATZWMERTTQALRGWVLDKSM3/action/replication_record"}},"created_at":"2026-07-05T11:14:25.372040+00:00","updated_at":"2026-07-05T11:14:25.372040+00:00"}