{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:EOYZZ3M6AQLAYZHJRZ3MT4CNIN","short_pith_number":"pith:EOYZZ3M6","schema_version":"1.0","canonical_sha256":"23b19ced9e04160c64e98e76c9f04d43652e6804c9823cc0687e3218a4d8bf76","source":{"kind":"arxiv","id":"2206.01197","version":1},"attestation_state":"computed","paper":{"title":"Hard Negative Sampling Strategies for Contrastive Representation Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Afrina Tabassum, Hoda Eldardiry, Ismini Lourentzou, Muntasir Wahed","submitted_at":"2022-06-02T17:55:15Z","abstract_excerpt":"One of the challenges in contrastive learning is the selection of appropriate \\textit{hard negative} examples, in the absence of label information. Random sampling or importance sampling methods based on feature similarity often lead to sub-optimal performance. In this work, we introduce UnReMix, a hard negative sampling strategy that takes into account anchor similarity, model uncertainty and representativeness. Experimental results on several benchmarks show that UnReMix improves negative sample selection, and subsequently downstream performance when compared to state-of-the-art contrastive "},"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":"2206.01197","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2022-06-02T17:55:15Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"eae56e3c8299aeb3d36c0c9537248c28b96083ee1d3a3f17e77a877b9ae8831d","abstract_canon_sha256":"682a63dd54ad58f4f1d61d261d531347b03936c2891b473e9267c4d38dd977be"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:28:37.895962Z","signature_b64":"yHOQwsdQ+D3OagGyC4jz7H6wjtpgop3oifDhlAhTabaVbwg+XPVnOBKFM+KBCprRCvsK5SEL+1juyJdr3ZF+CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"23b19ced9e04160c64e98e76c9f04d43652e6804c9823cc0687e3218a4d8bf76","last_reissued_at":"2026-07-05T04:28:37.895516Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:28:37.895516Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hard Negative Sampling Strategies for Contrastive Representation Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Afrina Tabassum, Hoda Eldardiry, Ismini Lourentzou, Muntasir Wahed","submitted_at":"2022-06-02T17:55:15Z","abstract_excerpt":"One of the challenges in contrastive learning is the selection of appropriate \\textit{hard negative} examples, in the absence of label information. Random sampling or importance sampling methods based on feature similarity often lead to sub-optimal performance. In this work, we introduce UnReMix, a hard negative sampling strategy that takes into account anchor similarity, model uncertainty and representativeness. Experimental results on several benchmarks show that UnReMix improves negative sample selection, and subsequently downstream performance when compared to state-of-the-art contrastive "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2206.01197","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/2206.01197/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":"2206.01197","created_at":"2026-07-05T04:28:37.895572+00:00"},{"alias_kind":"arxiv_version","alias_value":"2206.01197v1","created_at":"2026-07-05T04:28:37.895572+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2206.01197","created_at":"2026-07-05T04:28:37.895572+00:00"},{"alias_kind":"pith_short_12","alias_value":"EOYZZ3M6AQLA","created_at":"2026-07-05T04:28:37.895572+00:00"},{"alias_kind":"pith_short_16","alias_value":"EOYZZ3M6AQLAYZHJ","created_at":"2026-07-05T04:28:37.895572+00:00"},{"alias_kind":"pith_short_8","alias_value":"EOYZZ3M6","created_at":"2026-07-05T04:28:37.895572+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.11291","citing_title":"Optimal Representations for Generalized Contrastive Learning with Imbalanced Datasets","ref_index":49,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06157","citing_title":"HNC: Leveraging Hard Negative Captions towards Models with Fine-Grained Visual-Linguistic Comprehension Capabilities","ref_index":26,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/EOYZZ3M6AQLAYZHJRZ3MT4CNIN","json":"https://pith.science/pith/EOYZZ3M6AQLAYZHJRZ3MT4CNIN.json","graph_json":"https://pith.science/api/pith-number/EOYZZ3M6AQLAYZHJRZ3MT4CNIN/graph.json","events_json":"https://pith.science/api/pith-number/EOYZZ3M6AQLAYZHJRZ3MT4CNIN/events.json","paper":"https://pith.science/paper/EOYZZ3M6"},"agent_actions":{"view_html":"https://pith.science/pith/EOYZZ3M6AQLAYZHJRZ3MT4CNIN","download_json":"https://pith.science/pith/EOYZZ3M6AQLAYZHJRZ3MT4CNIN.json","view_paper":"https://pith.science/paper/EOYZZ3M6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2206.01197&json=true","fetch_graph":"https://pith.science/api/pith-number/EOYZZ3M6AQLAYZHJRZ3MT4CNIN/graph.json","fetch_events":"https://pith.science/api/pith-number/EOYZZ3M6AQLAYZHJRZ3MT4CNIN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EOYZZ3M6AQLAYZHJRZ3MT4CNIN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EOYZZ3M6AQLAYZHJRZ3MT4CNIN/action/storage_attestation","attest_author":"https://pith.science/pith/EOYZZ3M6AQLAYZHJRZ3MT4CNIN/action/author_attestation","sign_citation":"https://pith.science/pith/EOYZZ3M6AQLAYZHJRZ3MT4CNIN/action/citation_signature","submit_replication":"https://pith.science/pith/EOYZZ3M6AQLAYZHJRZ3MT4CNIN/action/replication_record"}},"created_at":"2026-07-05T04:28:37.895572+00:00","updated_at":"2026-07-05T04:28:37.895572+00:00"}