{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:6DYAXO3ODFG2WNF2GNDWACUDOH","short_pith_number":"pith:6DYAXO3O","schema_version":"1.0","canonical_sha256":"f0f00bbb6e194dab34ba3347600a8371c018496d3fa065b78187f14513e3e308","source":{"kind":"arxiv","id":"2305.02299","version":4},"attestation_state":"computed","paper":{"title":"Dynamic Sparse Training with Structured Sparsity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Anna Golubeva, Mihai Nica, Mike Lasby, Utku Evci, Yani Ioannou","submitted_at":"2023-05-03T17:48:55Z","abstract_excerpt":"Dynamic Sparse Training (DST) methods achieve state-of-the-art results in sparse neural network training, matching the generalization of dense models while enabling sparse training and inference. Although the resulting models are highly sparse and theoretically less computationally expensive, achieving speedups with unstructured sparsity on real-world hardware is challenging. In this work, we propose a sparse-to-sparse DST method, Structured RigL (SRigL), to learn a variant of fine-grained structured N:M sparsity by imposing a constant fan-in constraint. Using our empirical analysis of existin"},"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":"2305.02299","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-05-03T17:48:55Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"02c64728fc3f25da4810a47517c745fe4fad1f7a9e02cf666a8c4ccf55701cf1","abstract_canon_sha256":"2c6012093d36aaa06cc673dfbfddfd67f49844a928d9e050d80ab42459ef44d3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:47:56.552459Z","signature_b64":"oZN2ieDJSrqPbCFLQM5pRsakRE8ermV0TWCg3K5ztYcDWBitw2U3ED0FJEcchlOxw9A5qoQTdWpPbO3MeZA/AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f0f00bbb6e194dab34ba3347600a8371c018496d3fa065b78187f14513e3e308","last_reissued_at":"2026-07-05T07:47:56.551824Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:47:56.551824Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Dynamic Sparse Training with Structured Sparsity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Anna Golubeva, Mihai Nica, Mike Lasby, Utku Evci, Yani Ioannou","submitted_at":"2023-05-03T17:48:55Z","abstract_excerpt":"Dynamic Sparse Training (DST) methods achieve state-of-the-art results in sparse neural network training, matching the generalization of dense models while enabling sparse training and inference. Although the resulting models are highly sparse and theoretically less computationally expensive, achieving speedups with unstructured sparsity on real-world hardware is challenging. In this work, we propose a sparse-to-sparse DST method, Structured RigL (SRigL), to learn a variant of fine-grained structured N:M sparsity by imposing a constant fan-in constraint. Using our empirical analysis of existin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.02299","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/2305.02299/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":"2305.02299","created_at":"2026-07-05T07:47:56.551907+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.02299v4","created_at":"2026-07-05T07:47:56.551907+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.02299","created_at":"2026-07-05T07:47:56.551907+00:00"},{"alias_kind":"pith_short_12","alias_value":"6DYAXO3ODFG2","created_at":"2026-07-05T07:47:56.551907+00:00"},{"alias_kind":"pith_short_16","alias_value":"6DYAXO3ODFG2WNF2","created_at":"2026-07-05T07:47:56.551907+00:00"},{"alias_kind":"pith_short_8","alias_value":"6DYAXO3O","created_at":"2026-07-05T07:47:56.551907+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18923","citing_title":"GrapNet: A Programmable Dynamic-Architecture Neural Graph Substrate","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15435","citing_title":"On the Stability of Growth in Structural Plasticity","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21104","citing_title":"HORST: Composing Optimizer Geometries for Sparse Transformer Training","ref_index":148,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15435","citing_title":"On the Stability of Growth in Structural Plasticity","ref_index":24,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6DYAXO3ODFG2WNF2GNDWACUDOH","json":"https://pith.science/pith/6DYAXO3ODFG2WNF2GNDWACUDOH.json","graph_json":"https://pith.science/api/pith-number/6DYAXO3ODFG2WNF2GNDWACUDOH/graph.json","events_json":"https://pith.science/api/pith-number/6DYAXO3ODFG2WNF2GNDWACUDOH/events.json","paper":"https://pith.science/paper/6DYAXO3O"},"agent_actions":{"view_html":"https://pith.science/pith/6DYAXO3ODFG2WNF2GNDWACUDOH","download_json":"https://pith.science/pith/6DYAXO3ODFG2WNF2GNDWACUDOH.json","view_paper":"https://pith.science/paper/6DYAXO3O","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.02299&json=true","fetch_graph":"https://pith.science/api/pith-number/6DYAXO3ODFG2WNF2GNDWACUDOH/graph.json","fetch_events":"https://pith.science/api/pith-number/6DYAXO3ODFG2WNF2GNDWACUDOH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6DYAXO3ODFG2WNF2GNDWACUDOH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6DYAXO3ODFG2WNF2GNDWACUDOH/action/storage_attestation","attest_author":"https://pith.science/pith/6DYAXO3ODFG2WNF2GNDWACUDOH/action/author_attestation","sign_citation":"https://pith.science/pith/6DYAXO3ODFG2WNF2GNDWACUDOH/action/citation_signature","submit_replication":"https://pith.science/pith/6DYAXO3ODFG2WNF2GNDWACUDOH/action/replication_record"}},"created_at":"2026-07-05T07:47:56.551907+00:00","updated_at":"2026-07-05T07:47:56.551907+00:00"}