{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:VDZMI7AR5HK7QTOFEUBGYJP7MX","short_pith_number":"pith:VDZMI7AR","schema_version":"1.0","canonical_sha256":"a8f2c47c11e9d5f84dc525026c25ff65de662de03b02241157eeb433bf8c63e7","source":{"kind":"arxiv","id":"2210.06441","version":2},"attestation_state":"computed","paper":{"title":"How Much Data Are Augmentations Worth? An Investigation into Scaling Laws, Invariance, and Implicit Regularization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Andrew Gordon Wilson, Gowthami Somepalli, Jonas Geiping, Micah Goldblum, Ravid Shwartz-Ziv, Tom Goldstein","submitted_at":"2022-10-12T17:42:01Z","abstract_excerpt":"Despite the clear performance benefits of data augmentations, little is known about why they are so effective. In this paper, we disentangle several key mechanisms through which data augmentations operate. Establishing an exchange rate between augmented and additional real data, we find that in out-of-distribution testing scenarios, augmentations which yield samples that are diverse, but inconsistent with the data distribution can be even more valuable than additional training data. Moreover, we find that data augmentations which encourage invariances can be more valuable than invariance alone"},"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":"2210.06441","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2022-10-12T17:42:01Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"d0fe240e147b11ebafe614034a4dd38763006ec325025c61b6b19ee258e04242","abstract_canon_sha256":"729ad0e07ce64aaa5691d2f6292800cd4ecbc244ddaadb460bb00fb76cb004de"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:56:39.845077Z","signature_b64":"VvUZqbP0TBCTA1P00APeJDlaCAhs1VqshaRSRlRGuT2RUYo5BAtSvynsukLronrGwB0P9fH7o0kr87Yv+ElyCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a8f2c47c11e9d5f84dc525026c25ff65de662de03b02241157eeb433bf8c63e7","last_reissued_at":"2026-07-05T05:56:39.844671Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:56:39.844671Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Much Data Are Augmentations Worth? An Investigation into Scaling Laws, Invariance, and Implicit Regularization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.LG","authors_text":"Andrew Gordon Wilson, Gowthami Somepalli, Jonas Geiping, Micah Goldblum, Ravid Shwartz-Ziv, Tom Goldstein","submitted_at":"2022-10-12T17:42:01Z","abstract_excerpt":"Despite the clear performance benefits of data augmentations, little is known about why they are so effective. In this paper, we disentangle several key mechanisms through which data augmentations operate. Establishing an exchange rate between augmented and additional real data, we find that in out-of-distribution testing scenarios, augmentations which yield samples that are diverse, but inconsistent with the data distribution can be even more valuable than additional training data. Moreover, we find that data augmentations which encourage invariances can be more valuable than invariance alone"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2210.06441","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/2210.06441/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":"2210.06441","created_at":"2026-07-05T05:56:39.844727+00:00"},{"alias_kind":"arxiv_version","alias_value":"2210.06441v2","created_at":"2026-07-05T05:56:39.844727+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2210.06441","created_at":"2026-07-05T05:56:39.844727+00:00"},{"alias_kind":"pith_short_12","alias_value":"VDZMI7AR5HK7","created_at":"2026-07-05T05:56:39.844727+00:00"},{"alias_kind":"pith_short_16","alias_value":"VDZMI7AR5HK7QTOF","created_at":"2026-07-05T05:56:39.844727+00:00"},{"alias_kind":"pith_short_8","alias_value":"VDZMI7AR","created_at":"2026-07-05T05:56:39.844727+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2509.24244","citing_title":"Model Merging Scaling Laws in Large Language Models","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VDZMI7AR5HK7QTOFEUBGYJP7MX","json":"https://pith.science/pith/VDZMI7AR5HK7QTOFEUBGYJP7MX.json","graph_json":"https://pith.science/api/pith-number/VDZMI7AR5HK7QTOFEUBGYJP7MX/graph.json","events_json":"https://pith.science/api/pith-number/VDZMI7AR5HK7QTOFEUBGYJP7MX/events.json","paper":"https://pith.science/paper/VDZMI7AR"},"agent_actions":{"view_html":"https://pith.science/pith/VDZMI7AR5HK7QTOFEUBGYJP7MX","download_json":"https://pith.science/pith/VDZMI7AR5HK7QTOFEUBGYJP7MX.json","view_paper":"https://pith.science/paper/VDZMI7AR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2210.06441&json=true","fetch_graph":"https://pith.science/api/pith-number/VDZMI7AR5HK7QTOFEUBGYJP7MX/graph.json","fetch_events":"https://pith.science/api/pith-number/VDZMI7AR5HK7QTOFEUBGYJP7MX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VDZMI7AR5HK7QTOFEUBGYJP7MX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VDZMI7AR5HK7QTOFEUBGYJP7MX/action/storage_attestation","attest_author":"https://pith.science/pith/VDZMI7AR5HK7QTOFEUBGYJP7MX/action/author_attestation","sign_citation":"https://pith.science/pith/VDZMI7AR5HK7QTOFEUBGYJP7MX/action/citation_signature","submit_replication":"https://pith.science/pith/VDZMI7AR5HK7QTOFEUBGYJP7MX/action/replication_record"}},"created_at":"2026-07-05T05:56:39.844727+00:00","updated_at":"2026-07-05T05:56:39.844727+00:00"}