{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:UHCWB2HRGOKY3T236OJW4GWHKE","short_pith_number":"pith:UHCWB2HR","schema_version":"1.0","canonical_sha256":"a1c560e8f133958dcf5bf3936e1ac75136ea4c1318baa8525f44c36e2ac3eb11","source":{"kind":"arxiv","id":"2106.09291","version":1},"attestation_state":"computed","paper":{"title":"Towards Understanding Deep Learning from Noisy Labels with Small-Loss Criterion","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Wei Wang, Xian-Jin Gui, Zhang-Hao Tian","submitted_at":"2021-06-17T07:53:12Z","abstract_excerpt":"Deep neural networks need large amounts of labeled data to achieve good performance. In real-world applications, labels are usually collected from non-experts such as crowdsourcing to save cost and thus are noisy. In the past few years, deep learning methods for dealing with noisy labels have been developed, many of which are based on the small-loss criterion. However, there are few theoretical analyses to explain why these methods could learn well from noisy labels. In this paper, we theoretically explain why the widely-used small-loss criterion works. Based on the explanation, we reformalize"},"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":"2106.09291","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-17T07:53:12Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"1756f4af914260e6a7ff8e5cda391de467422969a4274d97ca5e62813dae57d7","abstract_canon_sha256":"9b595ab6f80316c9781eb09ace61127a2c63e9ca99823018c595981821c09e28"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:50:14.640048Z","signature_b64":"On45eUSL+hlOlGRKXYWBbEUubZapIdcO8Wn/lfrA6HxTGAolJIsD9Z0opP0fBUZykppd2so/FaOh3wGOQI1eAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a1c560e8f133958dcf5bf3936e1ac75136ea4c1318baa8525f44c36e2ac3eb11","last_reissued_at":"2026-07-05T02:50:14.639626Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:50:14.639626Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Towards Understanding Deep Learning from Noisy Labels with Small-Loss Criterion","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Wei Wang, Xian-Jin Gui, Zhang-Hao Tian","submitted_at":"2021-06-17T07:53:12Z","abstract_excerpt":"Deep neural networks need large amounts of labeled data to achieve good performance. In real-world applications, labels are usually collected from non-experts such as crowdsourcing to save cost and thus are noisy. In the past few years, deep learning methods for dealing with noisy labels have been developed, many of which are based on the small-loss criterion. However, there are few theoretical analyses to explain why these methods could learn well from noisy labels. In this paper, we theoretically explain why the widely-used small-loss criterion works. Based on the explanation, we reformalize"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.09291","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/2106.09291/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":"2106.09291","created_at":"2026-07-05T02:50:14.639681+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.09291v1","created_at":"2026-07-05T02:50:14.639681+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.09291","created_at":"2026-07-05T02:50:14.639681+00:00"},{"alias_kind":"pith_short_12","alias_value":"UHCWB2HRGOKY","created_at":"2026-07-05T02:50:14.639681+00:00"},{"alias_kind":"pith_short_16","alias_value":"UHCWB2HRGOKY3T23","created_at":"2026-07-05T02:50:14.639681+00:00"},{"alias_kind":"pith_short_8","alias_value":"UHCWB2HR","created_at":"2026-07-05T02:50:14.639681+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11583","citing_title":"Beyond the Golden Teacher: Enhancing Graph Learning through LLM-GNN Co-teaching","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20725","citing_title":"Holistic Reliability Propagation: Decoupling Annotation and Prediction for Robust Noisy-Label","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20727","citing_title":"GAMR: Geometric-Aware Manifold Regularization with Virtual Outlier Synthesis for Learning with Noisy Labels","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03993","citing_title":"Can LLMs Learn to Reason Robustly under Noisy Supervision?","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16562","citing_title":"See Through the Noise: Improving Domain Generalization in Gaze Estimation","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UHCWB2HRGOKY3T236OJW4GWHKE","json":"https://pith.science/pith/UHCWB2HRGOKY3T236OJW4GWHKE.json","graph_json":"https://pith.science/api/pith-number/UHCWB2HRGOKY3T236OJW4GWHKE/graph.json","events_json":"https://pith.science/api/pith-number/UHCWB2HRGOKY3T236OJW4GWHKE/events.json","paper":"https://pith.science/paper/UHCWB2HR"},"agent_actions":{"view_html":"https://pith.science/pith/UHCWB2HRGOKY3T236OJW4GWHKE","download_json":"https://pith.science/pith/UHCWB2HRGOKY3T236OJW4GWHKE.json","view_paper":"https://pith.science/paper/UHCWB2HR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.09291&json=true","fetch_graph":"https://pith.science/api/pith-number/UHCWB2HRGOKY3T236OJW4GWHKE/graph.json","fetch_events":"https://pith.science/api/pith-number/UHCWB2HRGOKY3T236OJW4GWHKE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UHCWB2HRGOKY3T236OJW4GWHKE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UHCWB2HRGOKY3T236OJW4GWHKE/action/storage_attestation","attest_author":"https://pith.science/pith/UHCWB2HRGOKY3T236OJW4GWHKE/action/author_attestation","sign_citation":"https://pith.science/pith/UHCWB2HRGOKY3T236OJW4GWHKE/action/citation_signature","submit_replication":"https://pith.science/pith/UHCWB2HRGOKY3T236OJW4GWHKE/action/replication_record"}},"created_at":"2026-07-05T02:50:14.639681+00:00","updated_at":"2026-07-05T02:50:14.639681+00:00"}