{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:45M2N62OPEAO6OJHA2TMFH3R2E","short_pith_number":"pith:45M2N62O","schema_version":"1.0","canonical_sha256":"e759a6fb4e7900ef392706a6c29f71d1192d8adb2e6ab2eef5ce025e7b60b3dc","source":{"kind":"arxiv","id":"2407.10969","version":3},"attestation_state":"computed","paper":{"title":"Q-Sparse: All Large Language Models can be Fully Sparsely-Activated","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Furu Wei, Hongyu Wang, Ruiping Wang, Shuming Ma","submitted_at":"2024-07-15T17:59:29Z","abstract_excerpt":"We introduce, Q-Sparse, a simple yet effective approach to training sparsely-activated large language models (LLMs). Q-Sparse enables full sparsity of activations in LLMs which can bring significant efficiency gains in inference. This is achieved by applying top-K sparsification to the activations and the straight-through-estimator to the training. We also introduce Block Q-Sparse for batch training and inference. The key results from this work are, (1) Q-Sparse can achieve results comparable to those of baseline LLMs while being much more efficient at inference time; (2) We present an inferen"},"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":"2407.10969","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-15T17:59:29Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"200f670f58b397f45de0035841a74a798b51469366f9491054a348de7e6c1398","abstract_canon_sha256":"fae982f4303cb4d29ce2311c788b3ab9cf7289c3c14f7e05582f7dbcc0e24a29"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:47:54.623672Z","signature_b64":"IqrFS/aUCC9dmTORvTOYF6uXXLmnXoZrm5BZNdmwqb8YJZH7MLscWliIWWvz5dvNwsGUsXebrcOeHp4GL1JOBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e759a6fb4e7900ef392706a6c29f71d1192d8adb2e6ab2eef5ce025e7b60b3dc","last_reissued_at":"2026-07-05T08:47:54.623213Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:47:54.623213Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Q-Sparse: All Large Language Models can be Fully Sparsely-Activated","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Furu Wei, Hongyu Wang, Ruiping Wang, Shuming Ma","submitted_at":"2024-07-15T17:59:29Z","abstract_excerpt":"We introduce, Q-Sparse, a simple yet effective approach to training sparsely-activated large language models (LLMs). Q-Sparse enables full sparsity of activations in LLMs which can bring significant efficiency gains in inference. This is achieved by applying top-K sparsification to the activations and the straight-through-estimator to the training. We also introduce Block Q-Sparse for batch training and inference. The key results from this work are, (1) Q-Sparse can achieve results comparable to those of baseline LLMs while being much more efficient at inference time; (2) We present an inferen"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.10969","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/2407.10969/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":"2407.10969","created_at":"2026-07-05T08:47:54.623274+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.10969v3","created_at":"2026-07-05T08:47:54.623274+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.10969","created_at":"2026-07-05T08:47:54.623274+00:00"},{"alias_kind":"pith_short_12","alias_value":"45M2N62OPEAO","created_at":"2026-07-05T08:47:54.623274+00:00"},{"alias_kind":"pith_short_16","alias_value":"45M2N62OPEAO6OJH","created_at":"2026-07-05T08:47:54.623274+00:00"},{"alias_kind":"pith_short_8","alias_value":"45M2N62O","created_at":"2026-07-05T08:47:54.623274+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.26632","citing_title":"RT-Lynx: Putting the GEMM Sparsity In a Right Way for Diffusion Models","ref_index":62,"is_internal_anchor":false},{"citing_arxiv_id":"2509.22166","citing_title":"Motivating Next-Gen Accelerators with Flexible (N:M) Activation Sparsity via Benchmarking Lightweight Post-Training Sparsification Approaches","ref_index":22,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/45M2N62OPEAO6OJHA2TMFH3R2E","json":"https://pith.science/pith/45M2N62OPEAO6OJHA2TMFH3R2E.json","graph_json":"https://pith.science/api/pith-number/45M2N62OPEAO6OJHA2TMFH3R2E/graph.json","events_json":"https://pith.science/api/pith-number/45M2N62OPEAO6OJHA2TMFH3R2E/events.json","paper":"https://pith.science/paper/45M2N62O"},"agent_actions":{"view_html":"https://pith.science/pith/45M2N62OPEAO6OJHA2TMFH3R2E","download_json":"https://pith.science/pith/45M2N62OPEAO6OJHA2TMFH3R2E.json","view_paper":"https://pith.science/paper/45M2N62O","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.10969&json=true","fetch_graph":"https://pith.science/api/pith-number/45M2N62OPEAO6OJHA2TMFH3R2E/graph.json","fetch_events":"https://pith.science/api/pith-number/45M2N62OPEAO6OJHA2TMFH3R2E/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/45M2N62OPEAO6OJHA2TMFH3R2E/action/timestamp_anchor","attest_storage":"https://pith.science/pith/45M2N62OPEAO6OJHA2TMFH3R2E/action/storage_attestation","attest_author":"https://pith.science/pith/45M2N62OPEAO6OJHA2TMFH3R2E/action/author_attestation","sign_citation":"https://pith.science/pith/45M2N62OPEAO6OJHA2TMFH3R2E/action/citation_signature","submit_replication":"https://pith.science/pith/45M2N62OPEAO6OJHA2TMFH3R2E/action/replication_record"}},"created_at":"2026-07-05T08:47:54.623274+00:00","updated_at":"2026-07-05T08:47:54.623274+00:00"}