{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:CJYMYPCMAINLTMLIH5PNQHBJMK","short_pith_number":"pith:CJYMYPCM","schema_version":"1.0","canonical_sha256":"1270cc3c4c021ab9b1683f5ed81c2962b02e9415c8b1bdb5e3ced3e60c52e00a","source":{"kind":"arxiv","id":"2110.08529","version":2},"attestation_state":"computed","paper":{"title":"Sharpness-Aware Minimization Improves Language Model Generalization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Dara Bahri, Hossein Mobahi, Yi Tay","submitted_at":"2021-10-16T09:44:06Z","abstract_excerpt":"The allure of superhuman-level capabilities has led to considerable interest in language models like GPT-3 and T5, wherein the research has, by and large, revolved around new model architectures, training tasks, and loss objectives, along with substantial engineering efforts to scale up model capacity and dataset size. Comparatively little work has been done to improve the generalization of these models through better optimization. In this work, we show that Sharpness-Aware Minimization (SAM), a recently proposed optimization procedure that encourages convergence to flatter minima, can substan"},"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":"2110.08529","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-10-16T09:44:06Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"f8b7a9aa0be9e0a4b0bd9b03fbdd25e9613431d7a3873d5a7255e59eb3f91109","abstract_canon_sha256":"2287a96f57392a67204821ef3e0103dbd343768b2448133e550a4b14d307c82a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:05:35.843959Z","signature_b64":"eCZnDObqzm78lYZRtcCHG3qtSAgyTq9f0TD5s+4+stNsojD7jc1XzNzkdFC+fQyA7UhMOSqQJiJ2TmkDK0ZvDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1270cc3c4c021ab9b1683f5ed81c2962b02e9415c8b1bdb5e3ced3e60c52e00a","last_reissued_at":"2026-07-05T04:05:35.843548Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:05:35.843548Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sharpness-Aware Minimization Improves Language Model Generalization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Dara Bahri, Hossein Mobahi, Yi Tay","submitted_at":"2021-10-16T09:44:06Z","abstract_excerpt":"The allure of superhuman-level capabilities has led to considerable interest in language models like GPT-3 and T5, wherein the research has, by and large, revolved around new model architectures, training tasks, and loss objectives, along with substantial engineering efforts to scale up model capacity and dataset size. Comparatively little work has been done to improve the generalization of these models through better optimization. In this work, we show that Sharpness-Aware Minimization (SAM), a recently proposed optimization procedure that encourages convergence to flatter minima, can substan"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2110.08529","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/2110.08529/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":"2110.08529","created_at":"2026-07-05T04:05:35.843604+00:00"},{"alias_kind":"arxiv_version","alias_value":"2110.08529v2","created_at":"2026-07-05T04:05:35.843604+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2110.08529","created_at":"2026-07-05T04:05:35.843604+00:00"},{"alias_kind":"pith_short_12","alias_value":"CJYMYPCMAINL","created_at":"2026-07-05T04:05:35.843604+00:00"},{"alias_kind":"pith_short_16","alias_value":"CJYMYPCMAINLTMLI","created_at":"2026-07-05T04:05:35.843604+00:00"},{"alias_kind":"pith_short_8","alias_value":"CJYMYPCM","created_at":"2026-07-05T04:05:35.843604+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05727","citing_title":"SAMPLe: SAM-based Optimizer for Prompt Learning in VLMs","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2602.07340","citing_title":"Revisiting Robustness for LLM Safety Alignment via Selective Geometry Control","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CJYMYPCMAINLTMLIH5PNQHBJMK","json":"https://pith.science/pith/CJYMYPCMAINLTMLIH5PNQHBJMK.json","graph_json":"https://pith.science/api/pith-number/CJYMYPCMAINLTMLIH5PNQHBJMK/graph.json","events_json":"https://pith.science/api/pith-number/CJYMYPCMAINLTMLIH5PNQHBJMK/events.json","paper":"https://pith.science/paper/CJYMYPCM"},"agent_actions":{"view_html":"https://pith.science/pith/CJYMYPCMAINLTMLIH5PNQHBJMK","download_json":"https://pith.science/pith/CJYMYPCMAINLTMLIH5PNQHBJMK.json","view_paper":"https://pith.science/paper/CJYMYPCM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2110.08529&json=true","fetch_graph":"https://pith.science/api/pith-number/CJYMYPCMAINLTMLIH5PNQHBJMK/graph.json","fetch_events":"https://pith.science/api/pith-number/CJYMYPCMAINLTMLIH5PNQHBJMK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CJYMYPCMAINLTMLIH5PNQHBJMK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CJYMYPCMAINLTMLIH5PNQHBJMK/action/storage_attestation","attest_author":"https://pith.science/pith/CJYMYPCMAINLTMLIH5PNQHBJMK/action/author_attestation","sign_citation":"https://pith.science/pith/CJYMYPCMAINLTMLIH5PNQHBJMK/action/citation_signature","submit_replication":"https://pith.science/pith/CJYMYPCMAINLTMLIH5PNQHBJMK/action/replication_record"}},"created_at":"2026-07-05T04:05:35.843604+00:00","updated_at":"2026-07-05T04:05:35.843604+00:00"}