{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:3TVD744YJZVAXFPT6QD3VYWJKA","short_pith_number":"pith:3TVD744Y","schema_version":"1.0","canonical_sha256":"dcea3ff3984e6a0b95f3f407bae2c9500e86c34d27eaa97c38284c01d061bdf4","source":{"kind":"arxiv","id":"2010.09875","version":2},"attestation_state":"computed","paper":{"title":"Combining Ensembles and Data Augmentation can Harm your Calibration","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Balaji Lakshminarayanan, Dustin Tran, Ghassen Jerfel, Jasper Snoek, Michael W. Dusenberry, Rafael Muller, Yeming Wen","submitted_at":"2020-10-19T21:25:22Z","abstract_excerpt":"Ensemble methods which average over multiple neural network predictions are a simple approach to improve a model's calibration and robustness. Similarly, data augmentation techniques, which encode prior information in the form of invariant feature transformations, are effective for improving calibration and robustness. In this paper, we show a surprising pathology: combining ensembles and data augmentation can harm model calibration. This leads to a trade-off in practice, whereby improved accuracy by combining the two techniques comes at the expense of calibration. On the other hand, selecting"},"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":"2010.09875","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-19T21:25:22Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"ae77dde23731cb60fb84109504a679ccf85a820cfd6f175d4da4226ae9cc27da","abstract_canon_sha256":"098d7c8d8fc99ab15f0a816f8ff80884d9dda7326552930e292504f8c6bff595"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:25:26.986079Z","signature_b64":"8hNik51l09R7qpzY8JX0s79tEh6cYuvawSzQRhjGsLgdjzvv/dp/PjR60E05jQDlgdjSpviclacIBANtyYWUAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dcea3ff3984e6a0b95f3f407bae2c9500e86c34d27eaa97c38284c01d061bdf4","last_reissued_at":"2026-07-05T02:25:26.985648Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:25:26.985648Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Combining Ensembles and Data Augmentation can Harm your Calibration","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Balaji Lakshminarayanan, Dustin Tran, Ghassen Jerfel, Jasper Snoek, Michael W. Dusenberry, Rafael Muller, Yeming Wen","submitted_at":"2020-10-19T21:25:22Z","abstract_excerpt":"Ensemble methods which average over multiple neural network predictions are a simple approach to improve a model's calibration and robustness. Similarly, data augmentation techniques, which encode prior information in the form of invariant feature transformations, are effective for improving calibration and robustness. In this paper, we show a surprising pathology: combining ensembles and data augmentation can harm model calibration. This leads to a trade-off in practice, whereby improved accuracy by combining the two techniques comes at the expense of calibration. On the other hand, selecting"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.09875","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/2010.09875/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":"2010.09875","created_at":"2026-07-05T02:25:26.985704+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.09875v2","created_at":"2026-07-05T02:25:26.985704+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.09875","created_at":"2026-07-05T02:25:26.985704+00:00"},{"alias_kind":"pith_short_12","alias_value":"3TVD744YJZVA","created_at":"2026-07-05T02:25:26.985704+00:00"},{"alias_kind":"pith_short_16","alias_value":"3TVD744YJZVAXFPT","created_at":"2026-07-05T02:25:26.985704+00:00"},{"alias_kind":"pith_short_8","alias_value":"3TVD744Y","created_at":"2026-07-05T02:25:26.985704+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2412.16963","citing_title":"LH-Mix: Local Hierarchy Correlation Guided Mixup over Hierarchical Prompt Tuning","ref_index":37,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3TVD744YJZVAXFPT6QD3VYWJKA","json":"https://pith.science/pith/3TVD744YJZVAXFPT6QD3VYWJKA.json","graph_json":"https://pith.science/api/pith-number/3TVD744YJZVAXFPT6QD3VYWJKA/graph.json","events_json":"https://pith.science/api/pith-number/3TVD744YJZVAXFPT6QD3VYWJKA/events.json","paper":"https://pith.science/paper/3TVD744Y"},"agent_actions":{"view_html":"https://pith.science/pith/3TVD744YJZVAXFPT6QD3VYWJKA","download_json":"https://pith.science/pith/3TVD744YJZVAXFPT6QD3VYWJKA.json","view_paper":"https://pith.science/paper/3TVD744Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.09875&json=true","fetch_graph":"https://pith.science/api/pith-number/3TVD744YJZVAXFPT6QD3VYWJKA/graph.json","fetch_events":"https://pith.science/api/pith-number/3TVD744YJZVAXFPT6QD3VYWJKA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3TVD744YJZVAXFPT6QD3VYWJKA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3TVD744YJZVAXFPT6QD3VYWJKA/action/storage_attestation","attest_author":"https://pith.science/pith/3TVD744YJZVAXFPT6QD3VYWJKA/action/author_attestation","sign_citation":"https://pith.science/pith/3TVD744YJZVAXFPT6QD3VYWJKA/action/citation_signature","submit_replication":"https://pith.science/pith/3TVD744YJZVAXFPT6QD3VYWJKA/action/replication_record"}},"created_at":"2026-07-05T02:25:26.985704+00:00","updated_at":"2026-07-05T02:25:26.985704+00:00"}