{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:NUYWIGSGAG5KML7YSQGDOAB3DY","short_pith_number":"pith:NUYWIGSG","schema_version":"1.0","canonical_sha256":"6d31641a4601baa62ff8940c37003b1e0bf720790f385ecf91a506cb1534b299","source":{"kind":"arxiv","id":"2110.05025","version":2},"attestation_state":"computed","paper":{"title":"Self-supervised Learning is More Robust to Dataset Imbalance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Adrien Gaidon, Hong Liu, Jeff Z. HaoChen, Tengyu Ma","submitted_at":"2021-10-11T06:29:56Z","abstract_excerpt":"Self-supervised learning (SSL) is a scalable way to learn general visual representations since it learns without labels. However, large-scale unlabeled datasets in the wild often have long-tailed label distributions, where we know little about the behavior of SSL. In this work, we systematically investigate self-supervised learning under dataset imbalance. First, we find out via extensive experiments that off-the-shelf self-supervised representations are already more robust to class imbalance than supervised representations. The performance gap between balanced and imbalanced pre-training with"},"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":"2110.05025","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2021-10-11T06:29:56Z","cross_cats_sorted":["cs.CV","stat.ML"],"title_canon_sha256":"81336ee7d57272480cbc9e23d206fd24cf31c0ae3a14d2f04deb9b8a63140ace","abstract_canon_sha256":"1e30079c49494259d182e1234d62e4834f4eb0c140bb5d5b2096310db15e90a4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:25:12.650480Z","signature_b64":"bpc1h1W1qgoOfUFwfbvojf6kNvo7MtW5f0oVyWH2iKULoA6xGr645CHBbsoDiYvHmBxe29ljxGl7kSi1o31ODQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6d31641a4601baa62ff8940c37003b1e0bf720790f385ecf91a506cb1534b299","last_reissued_at":"2026-07-05T04:25:12.649967Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:25:12.649967Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Self-supervised Learning is More Robust to Dataset Imbalance","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Adrien Gaidon, Hong Liu, Jeff Z. HaoChen, Tengyu Ma","submitted_at":"2021-10-11T06:29:56Z","abstract_excerpt":"Self-supervised learning (SSL) is a scalable way to learn general visual representations since it learns without labels. However, large-scale unlabeled datasets in the wild often have long-tailed label distributions, where we know little about the behavior of SSL. In this work, we systematically investigate self-supervised learning under dataset imbalance. First, we find out via extensive experiments that off-the-shelf self-supervised representations are already more robust to class imbalance than supervised representations. The performance gap between balanced and imbalanced pre-training with"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2110.05025","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/2110.05025/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":"2110.05025","created_at":"2026-07-05T04:25:12.650029+00:00"},{"alias_kind":"arxiv_version","alias_value":"2110.05025v2","created_at":"2026-07-05T04:25:12.650029+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2110.05025","created_at":"2026-07-05T04:25:12.650029+00:00"},{"alias_kind":"pith_short_12","alias_value":"NUYWIGSGAG5K","created_at":"2026-07-05T04:25:12.650029+00:00"},{"alias_kind":"pith_short_16","alias_value":"NUYWIGSGAG5KML7Y","created_at":"2026-07-05T04:25:12.650029+00:00"},{"alias_kind":"pith_short_8","alias_value":"NUYWIGSG","created_at":"2026-07-05T04:25:12.650029+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2505.13518","citing_title":"Data Balancing Strategies: A Systematic Survey of Resampling and Augmentation Methods","ref_index":123,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NUYWIGSGAG5KML7YSQGDOAB3DY","json":"https://pith.science/pith/NUYWIGSGAG5KML7YSQGDOAB3DY.json","graph_json":"https://pith.science/api/pith-number/NUYWIGSGAG5KML7YSQGDOAB3DY/graph.json","events_json":"https://pith.science/api/pith-number/NUYWIGSGAG5KML7YSQGDOAB3DY/events.json","paper":"https://pith.science/paper/NUYWIGSG"},"agent_actions":{"view_html":"https://pith.science/pith/NUYWIGSGAG5KML7YSQGDOAB3DY","download_json":"https://pith.science/pith/NUYWIGSGAG5KML7YSQGDOAB3DY.json","view_paper":"https://pith.science/paper/NUYWIGSG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2110.05025&json=true","fetch_graph":"https://pith.science/api/pith-number/NUYWIGSGAG5KML7YSQGDOAB3DY/graph.json","fetch_events":"https://pith.science/api/pith-number/NUYWIGSGAG5KML7YSQGDOAB3DY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NUYWIGSGAG5KML7YSQGDOAB3DY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NUYWIGSGAG5KML7YSQGDOAB3DY/action/storage_attestation","attest_author":"https://pith.science/pith/NUYWIGSGAG5KML7YSQGDOAB3DY/action/author_attestation","sign_citation":"https://pith.science/pith/NUYWIGSGAG5KML7YSQGDOAB3DY/action/citation_signature","submit_replication":"https://pith.science/pith/NUYWIGSGAG5KML7YSQGDOAB3DY/action/replication_record"}},"created_at":"2026-07-05T04:25:12.650029+00:00","updated_at":"2026-07-05T04:25:12.650029+00:00"}