{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:NEKSPQJMO4PNULTSXLCSE3JT5K","short_pith_number":"pith:NEKSPQJM","schema_version":"1.0","canonical_sha256":"691527c12c771eda2e72bac5226d33ea896a02873a7558234308f189090bd5c4","source":{"kind":"arxiv","id":"2605.27541","version":1},"attestation_state":"computed","paper":{"title":"SparseOpt: Addressing Normalization-induced Gradient Skew in Sparse Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ekansh Sharma, Mohammed Adnan, Rahul G. Krishnan, Rebekka Burkholz, Rohan Jain, Tom Jacobs, Yani Ioannou","submitted_at":"2026-05-26T18:14:19Z","abstract_excerpt":"Dynamic Sparse Training (DST) methods train neural networks by maintaining sparsity while dynamically adapting the network topology. Despite the promise of reduced computation, DST methods converge significantly slower than dense training, often requiring comparable training time to achieve similar accuracy. We demonstrate both analytically and empirically that Batch Normalization (BN) adversely affects sparse training, and propose SparseOpt, a sparsity-aware optimizer, to address this. Experiments on ResNet models across CIFAR-100 and ImageNet demonstrate consistently faster convergence and i"},"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":"2605.27541","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-26T18:14:19Z","cross_cats_sorted":[],"title_canon_sha256":"680233a45a331c052a19bda08bd8b18f408043882d696ae35c61ba86e9f1519b","abstract_canon_sha256":"54d7817751ac5a1b4102f3b98917e99a28f0d94787b3055e8ad3a8a83c6b0cc8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:14.888936Z","signature_b64":"qZli339avfCn3UaE8UqMlNi2Chc1/TavzUTyAZmxlT5fBB2aviZAApoFMuOc9ndCNuhBs4O7+zARshMQQyiIDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"691527c12c771eda2e72bac5226d33ea896a02873a7558234308f189090bd5c4","last_reissued_at":"2026-05-28T01:04:14.888247Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:14.888247Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SparseOpt: Addressing Normalization-induced Gradient Skew in Sparse Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ekansh Sharma, Mohammed Adnan, Rahul G. Krishnan, Rebekka Burkholz, Rohan Jain, Tom Jacobs, Yani Ioannou","submitted_at":"2026-05-26T18:14:19Z","abstract_excerpt":"Dynamic Sparse Training (DST) methods train neural networks by maintaining sparsity while dynamically adapting the network topology. Despite the promise of reduced computation, DST methods converge significantly slower than dense training, often requiring comparable training time to achieve similar accuracy. We demonstrate both analytically and empirically that Batch Normalization (BN) adversely affects sparse training, and propose SparseOpt, a sparsity-aware optimizer, to address this. Experiments on ResNet models across CIFAR-100 and ImageNet demonstrate consistently faster convergence and i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.27541","kind":"arxiv","version":1},"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/2605.27541/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":"2605.27541","created_at":"2026-05-28T01:04:14.888338+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.27541v1","created_at":"2026-05-28T01:04:14.888338+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.27541","created_at":"2026-05-28T01:04:14.888338+00:00"},{"alias_kind":"pith_short_12","alias_value":"NEKSPQJMO4PN","created_at":"2026-05-28T01:04:14.888338+00:00"},{"alias_kind":"pith_short_16","alias_value":"NEKSPQJMO4PNULTS","created_at":"2026-05-28T01:04:14.888338+00:00"},{"alias_kind":"pith_short_8","alias_value":"NEKSPQJM","created_at":"2026-05-28T01:04:14.888338+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NEKSPQJMO4PNULTSXLCSE3JT5K","json":"https://pith.science/pith/NEKSPQJMO4PNULTSXLCSE3JT5K.json","graph_json":"https://pith.science/api/pith-number/NEKSPQJMO4PNULTSXLCSE3JT5K/graph.json","events_json":"https://pith.science/api/pith-number/NEKSPQJMO4PNULTSXLCSE3JT5K/events.json","paper":"https://pith.science/paper/NEKSPQJM"},"agent_actions":{"view_html":"https://pith.science/pith/NEKSPQJMO4PNULTSXLCSE3JT5K","download_json":"https://pith.science/pith/NEKSPQJMO4PNULTSXLCSE3JT5K.json","view_paper":"https://pith.science/paper/NEKSPQJM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.27541&json=true","fetch_graph":"https://pith.science/api/pith-number/NEKSPQJMO4PNULTSXLCSE3JT5K/graph.json","fetch_events":"https://pith.science/api/pith-number/NEKSPQJMO4PNULTSXLCSE3JT5K/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NEKSPQJMO4PNULTSXLCSE3JT5K/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NEKSPQJMO4PNULTSXLCSE3JT5K/action/storage_attestation","attest_author":"https://pith.science/pith/NEKSPQJMO4PNULTSXLCSE3JT5K/action/author_attestation","sign_citation":"https://pith.science/pith/NEKSPQJMO4PNULTSXLCSE3JT5K/action/citation_signature","submit_replication":"https://pith.science/pith/NEKSPQJMO4PNULTSXLCSE3JT5K/action/replication_record"}},"created_at":"2026-05-28T01:04:14.888338+00:00","updated_at":"2026-05-28T01:04:14.888338+00:00"}