{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:MGKIW3DZULJGZQ6TF27C33TYVZ","short_pith_number":"pith:MGKIW3DZ","schema_version":"1.0","canonical_sha256":"61948b6c79a2d26cc3d32ebe2dee78ae6b91260e0e25891c736779ca8307fe58","source":{"kind":"arxiv","id":"2106.02866","version":2},"attestation_state":"computed","paper":{"title":"Conditional Contrastive Learning for Improving Fairness in Self-Supervised Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Han Zhao, Kun Zhang, Louis-Philippe Morency, Martin Q. Ma, Paul Pu Liang, Ruslan Salakhutdinov, Yao-Hung Hubert Tsai","submitted_at":"2021-06-05T10:51:26Z","abstract_excerpt":"Contrastive self-supervised learning (SSL) learns an embedding space that maps similar data pairs closer and dissimilar data pairs farther apart. Despite its success, one issue has been overlooked: the fairness aspect of representations learned using contrastive SSL. Without mitigation, contrastive SSL techniques can incorporate sensitive information such as gender or race and cause potentially unfair predictions on downstream tasks. In this paper, we propose a Conditional Contrastive Learning (CCL) approach to improve the fairness of contrastive SSL methods. Our approach samples positive and "},"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":"2106.02866","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-05T10:51:26Z","cross_cats_sorted":[],"title_canon_sha256":"eb52dde91e6a6ae1088429bb284adb0f6c11490d6aace22e386514a11870d33b","abstract_canon_sha256":"5404d894a21e2aebecb12b7fbb4b6830d534d58be7058161498b498ba26d85cc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:35:08.520740Z","signature_b64":"OQXYeWmsByWx9o2pIzaqbmLyZIRzoWUa8/K9I5xF+dRaAif/A0RrzM/dJ6MvX86fZmRj0v3zp0VqugjCqAooCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"61948b6c79a2d26cc3d32ebe2dee78ae6b91260e0e25891c736779ca8307fe58","last_reissued_at":"2026-07-05T04:35:08.520142Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:35:08.520142Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Conditional Contrastive Learning for Improving Fairness in Self-Supervised Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Han Zhao, Kun Zhang, Louis-Philippe Morency, Martin Q. Ma, Paul Pu Liang, Ruslan Salakhutdinov, Yao-Hung Hubert Tsai","submitted_at":"2021-06-05T10:51:26Z","abstract_excerpt":"Contrastive self-supervised learning (SSL) learns an embedding space that maps similar data pairs closer and dissimilar data pairs farther apart. Despite its success, one issue has been overlooked: the fairness aspect of representations learned using contrastive SSL. Without mitigation, contrastive SSL techniques can incorporate sensitive information such as gender or race and cause potentially unfair predictions on downstream tasks. In this paper, we propose a Conditional Contrastive Learning (CCL) approach to improve the fairness of contrastive SSL methods. Our approach samples positive and "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.02866","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/2106.02866/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":"2106.02866","created_at":"2026-07-05T04:35:08.520213+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.02866v2","created_at":"2026-07-05T04:35:08.520213+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.02866","created_at":"2026-07-05T04:35:08.520213+00:00"},{"alias_kind":"pith_short_12","alias_value":"MGKIW3DZULJG","created_at":"2026-07-05T04:35:08.520213+00:00"},{"alias_kind":"pith_short_16","alias_value":"MGKIW3DZULJGZQ6T","created_at":"2026-07-05T04:35:08.520213+00:00"},{"alias_kind":"pith_short_8","alias_value":"MGKIW3DZ","created_at":"2026-07-05T04:35:08.520213+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01392","citing_title":"Multi-Objective Exploration and Preference Optimization via Mutual Information","ref_index":111,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07770","citing_title":"Contrast encodes inductive bias: separating slow noise from dynamics in predictive representation learning","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2603.05582","citing_title":"Bias In, Bias Out? Finding Unbiased Subnetworks in Vanilla Models","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08512","citing_title":"MoMo: Conditioned Contrastive Representation Learning for Preference-Modulated Planning","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08512","citing_title":"MoMo: Conditioned Contrastive Representation Learning for Preference-Modulated Planning","ref_index":45,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MGKIW3DZULJGZQ6TF27C33TYVZ","json":"https://pith.science/pith/MGKIW3DZULJGZQ6TF27C33TYVZ.json","graph_json":"https://pith.science/api/pith-number/MGKIW3DZULJGZQ6TF27C33TYVZ/graph.json","events_json":"https://pith.science/api/pith-number/MGKIW3DZULJGZQ6TF27C33TYVZ/events.json","paper":"https://pith.science/paper/MGKIW3DZ"},"agent_actions":{"view_html":"https://pith.science/pith/MGKIW3DZULJGZQ6TF27C33TYVZ","download_json":"https://pith.science/pith/MGKIW3DZULJGZQ6TF27C33TYVZ.json","view_paper":"https://pith.science/paper/MGKIW3DZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.02866&json=true","fetch_graph":"https://pith.science/api/pith-number/MGKIW3DZULJGZQ6TF27C33TYVZ/graph.json","fetch_events":"https://pith.science/api/pith-number/MGKIW3DZULJGZQ6TF27C33TYVZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MGKIW3DZULJGZQ6TF27C33TYVZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MGKIW3DZULJGZQ6TF27C33TYVZ/action/storage_attestation","attest_author":"https://pith.science/pith/MGKIW3DZULJGZQ6TF27C33TYVZ/action/author_attestation","sign_citation":"https://pith.science/pith/MGKIW3DZULJGZQ6TF27C33TYVZ/action/citation_signature","submit_replication":"https://pith.science/pith/MGKIW3DZULJGZQ6TF27C33TYVZ/action/replication_record"}},"created_at":"2026-07-05T04:35:08.520213+00:00","updated_at":"2026-07-05T04:35:08.520213+00:00"}