{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FXIVIKNQGZ56W7W6VG6R2EYEOR","short_pith_number":"pith:FXIVIKNQ","schema_version":"1.0","canonical_sha256":"2dd15429b0367beb7edea9bd1d1304744dee2e6384cf2f3b051fce56b274a61e","source":{"kind":"arxiv","id":"2410.23672","version":1},"attestation_state":"computed","paper":{"title":"Provable Benefit of Cutout and CutMix for Feature Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Chulhee Yun, Junsoo Oh","submitted_at":"2024-10-31T06:41:10Z","abstract_excerpt":"Patch-level data augmentation techniques such as Cutout and CutMix have demonstrated significant efficacy in enhancing the performance of vision tasks. However, a comprehensive theoretical understanding of these methods remains elusive. In this paper, we study two-layer neural networks trained using three distinct methods: vanilla training without augmentation, Cutout training, and CutMix training. Our analysis focuses on a feature-noise data model, which consists of several label-dependent features of varying rarity and label-independent noises of differing strengths. Our theorems demonstrate"},"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":"2410.23672","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-31T06:41:10Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"6c13f83a9319d4f28f4c1cf35e773927862215a8dc652743f44fad00a82aa134","abstract_canon_sha256":"423815230579a9f9c712721edec69917092a93b600d0da4644899ddd8cf78329"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:29:03.695045Z","signature_b64":"62b/jcE8uYZ5GIiwKaJ/nPyrvhl+Qaecd8LemCaqbt4QaI2AGcwONMzhlju0loxKFOSOI8EagP7dZP2BL8fUBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2dd15429b0367beb7edea9bd1d1304744dee2e6384cf2f3b051fce56b274a61e","last_reissued_at":"2026-07-05T09:29:03.694549Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:29:03.694549Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Provable Benefit of Cutout and CutMix for Feature Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Chulhee Yun, Junsoo Oh","submitted_at":"2024-10-31T06:41:10Z","abstract_excerpt":"Patch-level data augmentation techniques such as Cutout and CutMix have demonstrated significant efficacy in enhancing the performance of vision tasks. However, a comprehensive theoretical understanding of these methods remains elusive. In this paper, we study two-layer neural networks trained using three distinct methods: vanilla training without augmentation, Cutout training, and CutMix training. Our analysis focuses on a feature-noise data model, which consists of several label-dependent features of varying rarity and label-independent noises of differing strengths. Our theorems demonstrate"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.23672","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/2410.23672/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":"2410.23672","created_at":"2026-07-05T09:29:03.694607+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.23672v1","created_at":"2026-07-05T09:29:03.694607+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.23672","created_at":"2026-07-05T09:29:03.694607+00:00"},{"alias_kind":"pith_short_12","alias_value":"FXIVIKNQGZ56","created_at":"2026-07-05T09:29:03.694607+00:00"},{"alias_kind":"pith_short_16","alias_value":"FXIVIKNQGZ56W7W6","created_at":"2026-07-05T09:29:03.694607+00:00"},{"alias_kind":"pith_short_8","alias_value":"FXIVIKNQ","created_at":"2026-07-05T09:29:03.694607+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.08940","citing_title":"Towards Understanding Why Data Augmentation Improves Generalization","ref_index":25,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FXIVIKNQGZ56W7W6VG6R2EYEOR","json":"https://pith.science/pith/FXIVIKNQGZ56W7W6VG6R2EYEOR.json","graph_json":"https://pith.science/api/pith-number/FXIVIKNQGZ56W7W6VG6R2EYEOR/graph.json","events_json":"https://pith.science/api/pith-number/FXIVIKNQGZ56W7W6VG6R2EYEOR/events.json","paper":"https://pith.science/paper/FXIVIKNQ"},"agent_actions":{"view_html":"https://pith.science/pith/FXIVIKNQGZ56W7W6VG6R2EYEOR","download_json":"https://pith.science/pith/FXIVIKNQGZ56W7W6VG6R2EYEOR.json","view_paper":"https://pith.science/paper/FXIVIKNQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.23672&json=true","fetch_graph":"https://pith.science/api/pith-number/FXIVIKNQGZ56W7W6VG6R2EYEOR/graph.json","fetch_events":"https://pith.science/api/pith-number/FXIVIKNQGZ56W7W6VG6R2EYEOR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FXIVIKNQGZ56W7W6VG6R2EYEOR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FXIVIKNQGZ56W7W6VG6R2EYEOR/action/storage_attestation","attest_author":"https://pith.science/pith/FXIVIKNQGZ56W7W6VG6R2EYEOR/action/author_attestation","sign_citation":"https://pith.science/pith/FXIVIKNQGZ56W7W6VG6R2EYEOR/action/citation_signature","submit_replication":"https://pith.science/pith/FXIVIKNQGZ56W7W6VG6R2EYEOR/action/replication_record"}},"created_at":"2026-07-05T09:29:03.694607+00:00","updated_at":"2026-07-05T09:29:03.694607+00:00"}