{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:SVDFBL5YBTIR3VZHKDIHIVD4F6","short_pith_number":"pith:SVDFBL5Y","schema_version":"1.0","canonical_sha256":"954650afb80cd11dd72750d074547c2fbd38ba2b37d176345697bb03a50a114a","source":{"kind":"arxiv","id":"2010.04592","version":2},"attestation_state":"computed","paper":{"title":"Contrastive Learning with Hard Negative Samples","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Ching-Yao Chuang, Joshua Robinson, Stefanie Jegelka, Suvrit Sra","submitted_at":"2020-10-09T14:18:53Z","abstract_excerpt":"How can you sample good negative examples for contrastive learning? We argue that, as with metric learning, contrastive learning of representations benefits from hard negative samples (i.e., points that are difficult to distinguish from an anchor point). The key challenge toward using hard negatives is that contrastive methods must remain unsupervised, making it infeasible to adopt existing negative sampling strategies that use true similarity information. In response, we develop a new family of unsupervised sampling methods for selecting hard negative samples where the user can control the ha"},"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":"2010.04592","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-09T14:18:53Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"5c31ccf8ad2ab7e3acb0e91016ec3402636cf073482adda60134b480d1db2a38","abstract_canon_sha256":"9b8c4971abc04db3212764a538dae458e82f1e69a9a3fe1ebce7ed8248edd467"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:09:13.334293Z","signature_b64":"Oiccd1iq5mGaEp0uNtiBTSlu+5kERr6qCPaRXX7+kjN2VQbIeeGBsbsdPUCdOiORnySqn6pvoDpssU1kFj+pAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"954650afb80cd11dd72750d074547c2fbd38ba2b37d176345697bb03a50a114a","last_reissued_at":"2026-07-05T02:09:13.333824Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:09:13.333824Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Contrastive Learning with Hard Negative Samples","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Ching-Yao Chuang, Joshua Robinson, Stefanie Jegelka, Suvrit Sra","submitted_at":"2020-10-09T14:18:53Z","abstract_excerpt":"How can you sample good negative examples for contrastive learning? We argue that, as with metric learning, contrastive learning of representations benefits from hard negative samples (i.e., points that are difficult to distinguish from an anchor point). The key challenge toward using hard negatives is that contrastive methods must remain unsupervised, making it infeasible to adopt existing negative sampling strategies that use true similarity information. In response, we develop a new family of unsupervised sampling methods for selecting hard negative samples where the user can control the ha"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.04592","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/2010.04592/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":"2010.04592","created_at":"2026-07-05T02:09:13.333879+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.04592v2","created_at":"2026-07-05T02:09:13.333879+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.04592","created_at":"2026-07-05T02:09:13.333879+00:00"},{"alias_kind":"pith_short_12","alias_value":"SVDFBL5YBTIR","created_at":"2026-07-05T02:09:13.333879+00:00"},{"alias_kind":"pith_short_16","alias_value":"SVDFBL5YBTIR3VZH","created_at":"2026-07-05T02:09:13.333879+00:00"},{"alias_kind":"pith_short_8","alias_value":"SVDFBL5Y","created_at":"2026-07-05T02:09:13.333879+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":18,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18852","citing_title":"Aligning Implied Statements for Implicit Hate Speech Generalizability with Context-Bounded Semi-hard Negative Mining","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2606.17298","citing_title":"Reasoning Text-to-Video Retrieval for Operating Room Clips via Action-Driven Digital Twins","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2606.15604","citing_title":"Parameter-Efficient Adaptation of SAM 3 for Automated ITV Generation from 4DCT Images","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02345","citing_title":"Doing well with less! On Sampling Techniques for Empirical Pairwise Loss Estimation/Minimization","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01334","citing_title":"HOLA: Holistic Multi-Modal Alignment for Open-Set 3D Recognition","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25598","citing_title":"SurfSurg6D: Geometry Consistent Dense Correspondence for Textureless Surgical Instrument Pose Estimation","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01079","citing_title":"Chameleon: Style-Content Disentangled Framework for Cross-Domain Object Compositing","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.01482","citing_title":"Beyond Topical Similarity: Contrastive Evidence Retrieval with Interpretable Attention Alignment in RAG","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2606.22481","citing_title":"Lighting-Consistent Object Transfer Across Radiance Fields","ref_index":150,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23744","citing_title":"Contrast to Detect: Dynamic Graph Contrastive Regularization for Unsupervised Anomaly Detection in Multivariate Time Series","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2411.18084","citing_title":"From Exploration to Revelation: Detecting Dark Patterns in Mobile Apps","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17854","citing_title":"Learning over Positive and Negative Edges with Contrastive Message Passing","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19752","citing_title":"MSAlign: Aligning Molecule and Mass Spectra Foundation Models for Metabolite Identification","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26057","citing_title":"Similarity Choice and Negative Scaling in Supervised Contrastive Learning for Deepfake Audio Detection","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10784","citing_title":"MASS-DPO: Multi-negative Active Sample Selection for Direct Policy Optimization","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25273","citing_title":"Combating Visual Neglect and Semantic Drift in Large Multimodal Models for Enhanced Cross-Modal Retrieval","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13313","citing_title":"Concrete Jungle: Towards Concreteness Paved Contrastive Negative Mining for Compositional Understanding","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12443","citing_title":"DiffusionPrint: Learning Generative Fingerprints for Diffusion-Based Inpainting Localization","ref_index":45,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SVDFBL5YBTIR3VZHKDIHIVD4F6","json":"https://pith.science/pith/SVDFBL5YBTIR3VZHKDIHIVD4F6.json","graph_json":"https://pith.science/api/pith-number/SVDFBL5YBTIR3VZHKDIHIVD4F6/graph.json","events_json":"https://pith.science/api/pith-number/SVDFBL5YBTIR3VZHKDIHIVD4F6/events.json","paper":"https://pith.science/paper/SVDFBL5Y"},"agent_actions":{"view_html":"https://pith.science/pith/SVDFBL5YBTIR3VZHKDIHIVD4F6","download_json":"https://pith.science/pith/SVDFBL5YBTIR3VZHKDIHIVD4F6.json","view_paper":"https://pith.science/paper/SVDFBL5Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.04592&json=true","fetch_graph":"https://pith.science/api/pith-number/SVDFBL5YBTIR3VZHKDIHIVD4F6/graph.json","fetch_events":"https://pith.science/api/pith-number/SVDFBL5YBTIR3VZHKDIHIVD4F6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SVDFBL5YBTIR3VZHKDIHIVD4F6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SVDFBL5YBTIR3VZHKDIHIVD4F6/action/storage_attestation","attest_author":"https://pith.science/pith/SVDFBL5YBTIR3VZHKDIHIVD4F6/action/author_attestation","sign_citation":"https://pith.science/pith/SVDFBL5YBTIR3VZHKDIHIVD4F6/action/citation_signature","submit_replication":"https://pith.science/pith/SVDFBL5YBTIR3VZHKDIHIVD4F6/action/replication_record"}},"created_at":"2026-07-05T02:09:13.333879+00:00","updated_at":"2026-07-05T02:09:13.333879+00:00"}