{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:2N2K5QWVIW3A52BTCQ5DLTT536","short_pith_number":"pith:2N2K5QWV","schema_version":"1.0","canonical_sha256":"d374aec2d545b60ee833143a35ce7ddfb2d5492320e6bee804ba56f3f88b67f1","source":{"kind":"arxiv","id":"2011.11156","version":2},"attestation_state":"computed","paper":{"title":"Better Aggregation in Test-Time Augmentation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Davis Blalock, Divya Shanmugam, Guha Balakrishnan, John Guttag","submitted_at":"2020-11-23T00:46:00Z","abstract_excerpt":"Test-time augmentation -- the aggregation of predictions across transformed versions of a test input -- is a common practice in image classification. Traditionally, predictions are combined using a simple average. In this paper, we present 1) experimental analyses that shed light on cases in which the simple average is suboptimal and 2) a method to address these shortcomings. A key finding is that even when test-time augmentation produces a net improvement in accuracy, it can change many correct predictions into incorrect predictions. We delve into when and why test-time augmentation changes a"},"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":"2011.11156","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2020-11-23T00:46:00Z","cross_cats_sorted":[],"title_canon_sha256":"4e473416c5e1f6cc2cdef7409fcbc08f6c7b6dd4994cc668e3d860223f0fb0cf","abstract_canon_sha256":"96e4fa7a8295a7b5675d7fef33ce1e6e30ecaa7d4aa9a6bfe77ab3f6b3997617"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:21:40.842858Z","signature_b64":"ycR0vpHAddmRyiHmYHR4uQbiG0eRFQcuI8Y/GK1ubRAHDqi8zbPf3QYMtuu1ZuNIyUFFEYXyAw3RA0sTAnpOBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d374aec2d545b60ee833143a35ce7ddfb2d5492320e6bee804ba56f3f88b67f1","last_reissued_at":"2026-07-05T03:21:40.842358Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:21:40.842358Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Better Aggregation in Test-Time Augmentation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Davis Blalock, Divya Shanmugam, Guha Balakrishnan, John Guttag","submitted_at":"2020-11-23T00:46:00Z","abstract_excerpt":"Test-time augmentation -- the aggregation of predictions across transformed versions of a test input -- is a common practice in image classification. Traditionally, predictions are combined using a simple average. In this paper, we present 1) experimental analyses that shed light on cases in which the simple average is suboptimal and 2) a method to address these shortcomings. A key finding is that even when test-time augmentation produces a net improvement in accuracy, it can change many correct predictions into incorrect predictions. We delve into when and why test-time augmentation changes a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2011.11156","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/2011.11156/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":"2011.11156","created_at":"2026-07-05T03:21:40.842418+00:00"},{"alias_kind":"arxiv_version","alias_value":"2011.11156v2","created_at":"2026-07-05T03:21:40.842418+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2011.11156","created_at":"2026-07-05T03:21:40.842418+00:00"},{"alias_kind":"pith_short_12","alias_value":"2N2K5QWVIW3A","created_at":"2026-07-05T03:21:40.842418+00:00"},{"alias_kind":"pith_short_16","alias_value":"2N2K5QWVIW3A52BT","created_at":"2026-07-05T03:21:40.842418+00:00"},{"alias_kind":"pith_short_8","alias_value":"2N2K5QWV","created_at":"2026-07-05T03:21:40.842418+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.18188","citing_title":"UTOPYA: A Multimodal Deep Learning Framework for Physics-Informed Anomaly Detection and Time-Series Prediction","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24347","citing_title":"Semantic Segmentation for Histopathology using Learned Regularization based on Global Proportions","ref_index":54,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16121","citing_title":"Beyond One-Size-Fits-All: Adaptive Test-Time Augmentation for Sequential Recommendation","ref_index":27,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2N2K5QWVIW3A52BTCQ5DLTT536","json":"https://pith.science/pith/2N2K5QWVIW3A52BTCQ5DLTT536.json","graph_json":"https://pith.science/api/pith-number/2N2K5QWVIW3A52BTCQ5DLTT536/graph.json","events_json":"https://pith.science/api/pith-number/2N2K5QWVIW3A52BTCQ5DLTT536/events.json","paper":"https://pith.science/paper/2N2K5QWV"},"agent_actions":{"view_html":"https://pith.science/pith/2N2K5QWVIW3A52BTCQ5DLTT536","download_json":"https://pith.science/pith/2N2K5QWVIW3A52BTCQ5DLTT536.json","view_paper":"https://pith.science/paper/2N2K5QWV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2011.11156&json=true","fetch_graph":"https://pith.science/api/pith-number/2N2K5QWVIW3A52BTCQ5DLTT536/graph.json","fetch_events":"https://pith.science/api/pith-number/2N2K5QWVIW3A52BTCQ5DLTT536/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2N2K5QWVIW3A52BTCQ5DLTT536/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2N2K5QWVIW3A52BTCQ5DLTT536/action/storage_attestation","attest_author":"https://pith.science/pith/2N2K5QWVIW3A52BTCQ5DLTT536/action/author_attestation","sign_citation":"https://pith.science/pith/2N2K5QWVIW3A52BTCQ5DLTT536/action/citation_signature","submit_replication":"https://pith.science/pith/2N2K5QWVIW3A52BTCQ5DLTT536/action/replication_record"}},"created_at":"2026-07-05T03:21:40.842418+00:00","updated_at":"2026-07-05T03:21:40.842418+00:00"}