{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:G5YKEWUEF5SEVUMEVAJR4TVIHU","short_pith_number":"pith:G5YKEWUE","schema_version":"1.0","canonical_sha256":"3770a25a842f644ad184a8131e4ea83d249a57ec9cb214e956b39232dbd8df1e","source":{"kind":"arxiv","id":"2409.17481","version":2},"attestation_state":"computed","paper":{"title":"MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Gongfan Fang, Greg Heinrich, Hongxu Yin, Jan Kautz, Jeff Pool, Pavlo Molchanov, Saurav Muralidharan, Xinchao Wang","submitted_at":"2024-09-26T02:37:41Z","abstract_excerpt":"Large Language Models (LLMs) are distinguished by their massive parameter counts, which typically result in significant redundancy. This work introduces MaskLLM, a learnable pruning method that establishes Semi-structured (or ``N:M'') Sparsity in LLMs, aimed at reducing computational overhead during inference. Instead of developing a new importance criterion, MaskLLM explicitly models N:M patterns as a learnable distribution through Gumbel Softmax sampling. This approach facilitates end-to-end training on large-scale datasets and offers two notable advantages: 1) High-quality Masks - our metho"},"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":"2409.17481","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-09-26T02:37:41Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"5cb0e558a0200435ecb0622bed62f5d7fdbaf92e567e89a2fad1f83357ba4211","abstract_canon_sha256":"4d6db4a1386bff9e2b514743a5297628ca56c278a0b7471e8b8e1f31c0160913"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:45:55.278875Z","signature_b64":"eQHvhYmNaz1pcrIz5SKANwm20Kmy0QQDasY8OoPLagOQdV55kfWUk2gCAo1zDV0m6ORK3HADvNRKsBtKBwx9Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3770a25a842f644ad184a8131e4ea83d249a57ec9cb214e956b39232dbd8df1e","last_reissued_at":"2026-07-05T09:45:55.278361Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:45:55.278361Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MaskLLM: Learnable Semi-Structured Sparsity for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Gongfan Fang, Greg Heinrich, Hongxu Yin, Jan Kautz, Jeff Pool, Pavlo Molchanov, Saurav Muralidharan, Xinchao Wang","submitted_at":"2024-09-26T02:37:41Z","abstract_excerpt":"Large Language Models (LLMs) are distinguished by their massive parameter counts, which typically result in significant redundancy. This work introduces MaskLLM, a learnable pruning method that establishes Semi-structured (or ``N:M'') Sparsity in LLMs, aimed at reducing computational overhead during inference. Instead of developing a new importance criterion, MaskLLM explicitly models N:M patterns as a learnable distribution through Gumbel Softmax sampling. This approach facilitates end-to-end training on large-scale datasets and offers two notable advantages: 1) High-quality Masks - our metho"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.17481","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/2409.17481/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":"2409.17481","created_at":"2026-07-05T09:45:55.278432+00:00"},{"alias_kind":"arxiv_version","alias_value":"2409.17481v2","created_at":"2026-07-05T09:45:55.278432+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.17481","created_at":"2026-07-05T09:45:55.278432+00:00"},{"alias_kind":"pith_short_12","alias_value":"G5YKEWUEF5SE","created_at":"2026-07-05T09:45:55.278432+00:00"},{"alias_kind":"pith_short_16","alias_value":"G5YKEWUEF5SEVUME","created_at":"2026-07-05T09:45:55.278432+00:00"},{"alias_kind":"pith_short_8","alias_value":"G5YKEWUE","created_at":"2026-07-05T09:45:55.278432+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.09375","citing_title":"31.1 A 14.08-to-135.69Token/s ReRAM-on-Logic Stacked Outlier-Free Large-Language-Model Accelerator with Block-Clustered Weight-Compression and Adaptive Parallel-Speculative-Decoding","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06402","citing_title":"SparseForge: Efficient Semi-Structured LLM Sparsification via Annealing of Hessian-Guided Soft-Mask","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/G5YKEWUEF5SEVUMEVAJR4TVIHU","json":"https://pith.science/pith/G5YKEWUEF5SEVUMEVAJR4TVIHU.json","graph_json":"https://pith.science/api/pith-number/G5YKEWUEF5SEVUMEVAJR4TVIHU/graph.json","events_json":"https://pith.science/api/pith-number/G5YKEWUEF5SEVUMEVAJR4TVIHU/events.json","paper":"https://pith.science/paper/G5YKEWUE"},"agent_actions":{"view_html":"https://pith.science/pith/G5YKEWUEF5SEVUMEVAJR4TVIHU","download_json":"https://pith.science/pith/G5YKEWUEF5SEVUMEVAJR4TVIHU.json","view_paper":"https://pith.science/paper/G5YKEWUE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2409.17481&json=true","fetch_graph":"https://pith.science/api/pith-number/G5YKEWUEF5SEVUMEVAJR4TVIHU/graph.json","fetch_events":"https://pith.science/api/pith-number/G5YKEWUEF5SEVUMEVAJR4TVIHU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/G5YKEWUEF5SEVUMEVAJR4TVIHU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/G5YKEWUEF5SEVUMEVAJR4TVIHU/action/storage_attestation","attest_author":"https://pith.science/pith/G5YKEWUEF5SEVUMEVAJR4TVIHU/action/author_attestation","sign_citation":"https://pith.science/pith/G5YKEWUEF5SEVUMEVAJR4TVIHU/action/citation_signature","submit_replication":"https://pith.science/pith/G5YKEWUEF5SEVUMEVAJR4TVIHU/action/replication_record"}},"created_at":"2026-07-05T09:45:55.278432+00:00","updated_at":"2026-07-05T09:45:55.278432+00:00"}