{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:V2R6Q3K5LG4WTDXDKRIJHVG5NC","short_pith_number":"pith:V2R6Q3K5","schema_version":"1.0","canonical_sha256":"aea3e86d5d59b9698ee3545093d4dd68baf4b9f2f66625e955b533f5b50e5ade","source":{"kind":"arxiv","id":"2007.00224","version":3},"attestation_state":"computed","paper":{"title":"Debiased Contrastive Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Antonio Torralba, Ching-Yao Chuang, Joshua Robinson, Lin Yen-Chen, Stefanie Jegelka","submitted_at":"2020-07-01T04:25:24Z","abstract_excerpt":"A prominent technique for self-supervised representation learning has been to contrast semantically similar and dissimilar pairs of samples. Without access to labels, dissimilar (negative) points are typically taken to be randomly sampled datapoints, implicitly accepting that these points may, in reality, actually have the same label. Perhaps unsurprisingly, we observe that sampling negative examples from truly different labels improves performance, in a synthetic setting where labels are available. Motivated by this observation, we develop a debiased contrastive objective that corrects for th"},"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":"2007.00224","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-07-01T04:25:24Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"80633ac42c9207a90787f0dacd37a768aab9c5de042135085679b8c3b302a4bd","abstract_canon_sha256":"b038394bbc29c968e200a5b21ee5154523c4ab44c1c1ac8dda1d5894f7e4f569"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:44:49.293172Z","signature_b64":"ZhwoLZoQGiYc5qHmAc6koWN3V0MT3slbr8Rr7KTsXyXdBzLxoJs/nz6J7df/BwGm1uLivkTOSRIm6lG9Yb07Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"aea3e86d5d59b9698ee3545093d4dd68baf4b9f2f66625e955b533f5b50e5ade","last_reissued_at":"2026-07-05T01:44:49.292585Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:44:49.292585Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Debiased Contrastive Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Antonio Torralba, Ching-Yao Chuang, Joshua Robinson, Lin Yen-Chen, Stefanie Jegelka","submitted_at":"2020-07-01T04:25:24Z","abstract_excerpt":"A prominent technique for self-supervised representation learning has been to contrast semantically similar and dissimilar pairs of samples. Without access to labels, dissimilar (negative) points are typically taken to be randomly sampled datapoints, implicitly accepting that these points may, in reality, actually have the same label. Perhaps unsurprisingly, we observe that sampling negative examples from truly different labels improves performance, in a synthetic setting where labels are available. Motivated by this observation, we develop a debiased contrastive objective that corrects for th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2007.00224","kind":"arxiv","version":3},"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/2007.00224/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":"2007.00224","created_at":"2026-07-05T01:44:49.292659+00:00"},{"alias_kind":"arxiv_version","alias_value":"2007.00224v3","created_at":"2026-07-05T01:44:49.292659+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2007.00224","created_at":"2026-07-05T01:44:49.292659+00:00"},{"alias_kind":"pith_short_12","alias_value":"V2R6Q3K5LG4W","created_at":"2026-07-05T01:44:49.292659+00:00"},{"alias_kind":"pith_short_16","alias_value":"V2R6Q3K5LG4WTDXD","created_at":"2026-07-05T01:44:49.292659+00:00"},{"alias_kind":"pith_short_8","alias_value":"V2R6Q3K5","created_at":"2026-07-05T01:44:49.292659+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.06452","citing_title":"TRUST: Leveraging Text Robustness for Unsupervised Domain Adaptation","ref_index":11,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/V2R6Q3K5LG4WTDXDKRIJHVG5NC","json":"https://pith.science/pith/V2R6Q3K5LG4WTDXDKRIJHVG5NC.json","graph_json":"https://pith.science/api/pith-number/V2R6Q3K5LG4WTDXDKRIJHVG5NC/graph.json","events_json":"https://pith.science/api/pith-number/V2R6Q3K5LG4WTDXDKRIJHVG5NC/events.json","paper":"https://pith.science/paper/V2R6Q3K5"},"agent_actions":{"view_html":"https://pith.science/pith/V2R6Q3K5LG4WTDXDKRIJHVG5NC","download_json":"https://pith.science/pith/V2R6Q3K5LG4WTDXDKRIJHVG5NC.json","view_paper":"https://pith.science/paper/V2R6Q3K5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2007.00224&json=true","fetch_graph":"https://pith.science/api/pith-number/V2R6Q3K5LG4WTDXDKRIJHVG5NC/graph.json","fetch_events":"https://pith.science/api/pith-number/V2R6Q3K5LG4WTDXDKRIJHVG5NC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V2R6Q3K5LG4WTDXDKRIJHVG5NC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V2R6Q3K5LG4WTDXDKRIJHVG5NC/action/storage_attestation","attest_author":"https://pith.science/pith/V2R6Q3K5LG4WTDXDKRIJHVG5NC/action/author_attestation","sign_citation":"https://pith.science/pith/V2R6Q3K5LG4WTDXDKRIJHVG5NC/action/citation_signature","submit_replication":"https://pith.science/pith/V2R6Q3K5LG4WTDXDKRIJHVG5NC/action/replication_record"}},"created_at":"2026-07-05T01:44:49.292659+00:00","updated_at":"2026-07-05T01:44:49.292659+00:00"}