{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:UY3N5HBBJ3EMZARVS4KM7XQOL2","short_pith_number":"pith:UY3N5HBB","schema_version":"1.0","canonical_sha256":"a636de9c214ec8cc82359714cfde0e5ea5375b3f16fa71e0bf03bb192321115e","source":{"kind":"arxiv","id":"2110.02027","version":3},"attestation_state":"computed","paper":{"title":"ProGCL: Rethinking Hard Negative Mining in Graph Contrastive Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Ge Wang, Jintao Chen, Jun Xia, Lirong Wu, Stan Z.Li","submitted_at":"2021-10-05T13:15:59Z","abstract_excerpt":"Contrastive Learning (CL) has emerged as a dominant technique for unsupervised representation learning which embeds augmented versions of the anchor close to each other (positive samples) and pushes the embeddings of other samples (negatives) apart. As revealed in recent studies, CL can benefit from hard negatives (negatives that are most similar to the anchor). However, we observe limited benefits when we adopt existing hard negative mining techniques of other domains in Graph Contrastive Learning (GCL). We perform both experimental and theoretical analysis on this phenomenon and find it can "},"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":"2110.02027","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2021-10-05T13:15:59Z","cross_cats_sorted":[],"title_canon_sha256":"3f1f4d5a170631af885da27cd602820f1f562b1a3a908175ab09d3b75caeebdd","abstract_canon_sha256":"7e354097c562443522fe48bd36eecb9b05bb1313bbfefc48ea027cdb91f6c3b5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:31:25.325309Z","signature_b64":"0FSdlo2Rpf3FzAiWCGqqruv2gkqRSi8Lx0w79+STr5c72bw0RgITvFne7HpWsP2V0391sSEexpm50MUGS9zCAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a636de9c214ec8cc82359714cfde0e5ea5375b3f16fa71e0bf03bb192321115e","last_reissued_at":"2026-07-05T04:31:25.324879Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:31:25.324879Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ProGCL: Rethinking Hard Negative Mining in Graph Contrastive Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Ge Wang, Jintao Chen, Jun Xia, Lirong Wu, Stan Z.Li","submitted_at":"2021-10-05T13:15:59Z","abstract_excerpt":"Contrastive Learning (CL) has emerged as a dominant technique for unsupervised representation learning which embeds augmented versions of the anchor close to each other (positive samples) and pushes the embeddings of other samples (negatives) apart. As revealed in recent studies, CL can benefit from hard negatives (negatives that are most similar to the anchor). However, we observe limited benefits when we adopt existing hard negative mining techniques of other domains in Graph Contrastive Learning (GCL). We perform both experimental and theoretical analysis on this phenomenon and find it can "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2110.02027","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/2110.02027/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":"2110.02027","created_at":"2026-07-05T04:31:25.324939+00:00"},{"alias_kind":"arxiv_version","alias_value":"2110.02027v3","created_at":"2026-07-05T04:31:25.324939+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2110.02027","created_at":"2026-07-05T04:31:25.324939+00:00"},{"alias_kind":"pith_short_12","alias_value":"UY3N5HBBJ3EM","created_at":"2026-07-05T04:31:25.324939+00:00"},{"alias_kind":"pith_short_16","alias_value":"UY3N5HBBJ3EMZARV","created_at":"2026-07-05T04:31:25.324939+00:00"},{"alias_kind":"pith_short_8","alias_value":"UY3N5HBB","created_at":"2026-07-05T04:31:25.324939+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27747","citing_title":"UNICS: Multilingual Code Search via Unified Pseudocode and Contrastive Transfer Learning","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2510.00767","citing_title":"Color2Struct: efficient and accurate deep-learning inverse design of structural color with controllable inference","ref_index":96,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UY3N5HBBJ3EMZARVS4KM7XQOL2","json":"https://pith.science/pith/UY3N5HBBJ3EMZARVS4KM7XQOL2.json","graph_json":"https://pith.science/api/pith-number/UY3N5HBBJ3EMZARVS4KM7XQOL2/graph.json","events_json":"https://pith.science/api/pith-number/UY3N5HBBJ3EMZARVS4KM7XQOL2/events.json","paper":"https://pith.science/paper/UY3N5HBB"},"agent_actions":{"view_html":"https://pith.science/pith/UY3N5HBBJ3EMZARVS4KM7XQOL2","download_json":"https://pith.science/pith/UY3N5HBBJ3EMZARVS4KM7XQOL2.json","view_paper":"https://pith.science/paper/UY3N5HBB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2110.02027&json=true","fetch_graph":"https://pith.science/api/pith-number/UY3N5HBBJ3EMZARVS4KM7XQOL2/graph.json","fetch_events":"https://pith.science/api/pith-number/UY3N5HBBJ3EMZARVS4KM7XQOL2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UY3N5HBBJ3EMZARVS4KM7XQOL2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UY3N5HBBJ3EMZARVS4KM7XQOL2/action/storage_attestation","attest_author":"https://pith.science/pith/UY3N5HBBJ3EMZARVS4KM7XQOL2/action/author_attestation","sign_citation":"https://pith.science/pith/UY3N5HBBJ3EMZARVS4KM7XQOL2/action/citation_signature","submit_replication":"https://pith.science/pith/UY3N5HBBJ3EMZARVS4KM7XQOL2/action/replication_record"}},"created_at":"2026-07-05T04:31:25.324939+00:00","updated_at":"2026-07-05T04:31:25.324939+00:00"}