{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:3WH35FDEROEK4BNTQVXDOEY6AV","short_pith_number":"pith:3WH35FDE","schema_version":"1.0","canonical_sha256":"dd8fbe94648b88ae05b3856e37131e0548c24e59d6aeb5e9bf2506ab9cb2536c","source":{"kind":"arxiv","id":"2505.23866","version":2},"attestation_state":"computed","paper":{"title":"Towards Understanding The Calibration Benefits of Sharpness-Aware Minimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chengli Tan, Guang Dai, Haishan Ye, Jiangshe Zhang, Junmin Liu, Yong Xu, Yubo Zhou, Yunda Hao, Zengjie Song, Zixiang Zhao","submitted_at":"2025-05-29T09:55:29Z","abstract_excerpt":"Deep neural networks have been increasingly used in safety-critical applications such as medical diagnosis and autonomous driving. However, many studies suggest that they are prone to being poorly calibrated and have a propensity for overconfidence, which may have disastrous consequences. In this paper, unlike standard training such as stochastic gradient descent, we show that the recently proposed sharpness-aware minimization (SAM) counteracts this tendency towards overconfidence. The theoretical analysis suggests that SAM allows us to learn models that are already well-calibrated by implicit"},"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":"2505.23866","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-05-29T09:55:29Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e3343afa4456bf18e6be008958aba5c37016b642d6755224b71440b13caf3bf5","abstract_canon_sha256":"17d3b40ca421ad9e48086376b541955d8957f846c0df0019aea311e905b34904"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-25T01:17:44.798788Z","signature_b64":"aH2EDk0mau5FqDRF7zj7OfZt1MQKCPnwbWG4UGat/SThdKm0Awfxz2eSeoojNAgkVZAy7HeeVVY/48oD7ZznAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dd8fbe94648b88ae05b3856e37131e0548c24e59d6aeb5e9bf2506ab9cb2536c","last_reissued_at":"2026-06-25T01:17:44.798291Z","signature_status":"signed_v1","first_computed_at":"2026-06-25T01:17:44.798291Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Towards Understanding The Calibration Benefits of Sharpness-Aware Minimization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chengli Tan, Guang Dai, Haishan Ye, Jiangshe Zhang, Junmin Liu, Yong Xu, Yubo Zhou, Yunda Hao, Zengjie Song, Zixiang Zhao","submitted_at":"2025-05-29T09:55:29Z","abstract_excerpt":"Deep neural networks have been increasingly used in safety-critical applications such as medical diagnosis and autonomous driving. However, many studies suggest that they are prone to being poorly calibrated and have a propensity for overconfidence, which may have disastrous consequences. In this paper, unlike standard training such as stochastic gradient descent, we show that the recently proposed sharpness-aware minimization (SAM) counteracts this tendency towards overconfidence. The theoretical analysis suggests that SAM allows us to learn models that are already well-calibrated by implicit"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.23866","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/2505.23866/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":"2505.23866","created_at":"2026-06-25T01:17:44.798360+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.23866v2","created_at":"2026-06-25T01:17:44.798360+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.23866","created_at":"2026-06-25T01:17:44.798360+00:00"},{"alias_kind":"pith_short_12","alias_value":"3WH35FDEROEK","created_at":"2026-06-25T01:17:44.798360+00:00"},{"alias_kind":"pith_short_16","alias_value":"3WH35FDEROEK4BNT","created_at":"2026-06-25T01:17:44.798360+00:00"},{"alias_kind":"pith_short_8","alias_value":"3WH35FDE","created_at":"2026-06-25T01:17:44.798360+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.27715","citing_title":"Improving Calibration in Test-Time Prompt Tuning for Vision-Language Models via Data-Free Flatness-Aware Prompt Pretraining","ref_index":36,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3WH35FDEROEK4BNTQVXDOEY6AV","json":"https://pith.science/pith/3WH35FDEROEK4BNTQVXDOEY6AV.json","graph_json":"https://pith.science/api/pith-number/3WH35FDEROEK4BNTQVXDOEY6AV/graph.json","events_json":"https://pith.science/api/pith-number/3WH35FDEROEK4BNTQVXDOEY6AV/events.json","paper":"https://pith.science/paper/3WH35FDE"},"agent_actions":{"view_html":"https://pith.science/pith/3WH35FDEROEK4BNTQVXDOEY6AV","download_json":"https://pith.science/pith/3WH35FDEROEK4BNTQVXDOEY6AV.json","view_paper":"https://pith.science/paper/3WH35FDE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.23866&json=true","fetch_graph":"https://pith.science/api/pith-number/3WH35FDEROEK4BNTQVXDOEY6AV/graph.json","fetch_events":"https://pith.science/api/pith-number/3WH35FDEROEK4BNTQVXDOEY6AV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3WH35FDEROEK4BNTQVXDOEY6AV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3WH35FDEROEK4BNTQVXDOEY6AV/action/storage_attestation","attest_author":"https://pith.science/pith/3WH35FDEROEK4BNTQVXDOEY6AV/action/author_attestation","sign_citation":"https://pith.science/pith/3WH35FDEROEK4BNTQVXDOEY6AV/action/citation_signature","submit_replication":"https://pith.science/pith/3WH35FDEROEK4BNTQVXDOEY6AV/action/replication_record"}},"created_at":"2026-06-25T01:17:44.798360+00:00","updated_at":"2026-06-25T01:17:44.798360+00:00"}