{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:APSPQQHNCW63DMJPDA24FKZONK","short_pith_number":"pith:APSPQQHN","schema_version":"1.0","canonical_sha256":"03e4f840ed15bdb1b12f1835c2ab2e6a9a752826ce18a6ca96c16be651f43e59","source":{"kind":"arxiv","id":"2402.10150","version":1},"attestation_state":"computed","paper":{"title":"$f$-MICL: Understanding and Generalizing InfoNCE-based Contrastive Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Guojun Zhang, Hongyu Guo, Sun Sun, Yaoliang Yu, Yiwei Lu","submitted_at":"2024-02-15T17:57:54Z","abstract_excerpt":"In self-supervised contrastive learning, a widely-adopted objective function is InfoNCE, which uses the heuristic cosine similarity for the representation comparison, and is closely related to maximizing the Kullback-Leibler (KL)-based mutual information. In this paper, we aim at answering two intriguing questions: (1) Can we go beyond the KL-based objective? (2) Besides the popular cosine similarity, can we design a better similarity function? We provide answers to both questions by generalizing the KL-based mutual information to the $f$-Mutual Information in Contrastive Learning ($f$-MICL) u"},"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":"2402.10150","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-15T17:57:54Z","cross_cats_sorted":[],"title_canon_sha256":"9db11274ccee7bbdf877c1d6fd0b8a89ad8a7e8407c02bce0d4b5d8b068a662b","abstract_canon_sha256":"7df3ff22f7854acd04c47bbbd9c48fcefac19ebb2b8743adea99ef24488b98f1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:45:41.986559Z","signature_b64":"5s3SKf2gdtXOm1IC60Hab1/qkaaVTmqLR/o8875rJiehVTaMG2dHkbEu0EdCsu62c/MXQ9KKte8eTXU6nSDYCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"03e4f840ed15bdb1b12f1835c2ab2e6a9a752826ce18a6ca96c16be651f43e59","last_reissued_at":"2026-07-05T07:45:41.986172Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:45:41.986172Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"$f$-MICL: Understanding and Generalizing InfoNCE-based Contrastive Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Guojun Zhang, Hongyu Guo, Sun Sun, Yaoliang Yu, Yiwei Lu","submitted_at":"2024-02-15T17:57:54Z","abstract_excerpt":"In self-supervised contrastive learning, a widely-adopted objective function is InfoNCE, which uses the heuristic cosine similarity for the representation comparison, and is closely related to maximizing the Kullback-Leibler (KL)-based mutual information. In this paper, we aim at answering two intriguing questions: (1) Can we go beyond the KL-based objective? (2) Besides the popular cosine similarity, can we design a better similarity function? We provide answers to both questions by generalizing the KL-based mutual information to the $f$-Mutual Information in Contrastive Learning ($f$-MICL) u"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.10150","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/2402.10150/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":"2402.10150","created_at":"2026-07-05T07:45:41.986238+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.10150v1","created_at":"2026-07-05T07:45:41.986238+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.10150","created_at":"2026-07-05T07:45:41.986238+00:00"},{"alias_kind":"pith_short_12","alias_value":"APSPQQHNCW63","created_at":"2026-07-05T07:45:41.986238+00:00"},{"alias_kind":"pith_short_16","alias_value":"APSPQQHNCW63DMJP","created_at":"2026-07-05T07:45:41.986238+00:00"},{"alias_kind":"pith_short_8","alias_value":"APSPQQHN","created_at":"2026-07-05T07:45:41.986238+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.18671","citing_title":"Self-Supervised Evolution Operator Learning for High-Dimensional Dynamical Systems","ref_index":60,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/APSPQQHNCW63DMJPDA24FKZONK","json":"https://pith.science/pith/APSPQQHNCW63DMJPDA24FKZONK.json","graph_json":"https://pith.science/api/pith-number/APSPQQHNCW63DMJPDA24FKZONK/graph.json","events_json":"https://pith.science/api/pith-number/APSPQQHNCW63DMJPDA24FKZONK/events.json","paper":"https://pith.science/paper/APSPQQHN"},"agent_actions":{"view_html":"https://pith.science/pith/APSPQQHNCW63DMJPDA24FKZONK","download_json":"https://pith.science/pith/APSPQQHNCW63DMJPDA24FKZONK.json","view_paper":"https://pith.science/paper/APSPQQHN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.10150&json=true","fetch_graph":"https://pith.science/api/pith-number/APSPQQHNCW63DMJPDA24FKZONK/graph.json","fetch_events":"https://pith.science/api/pith-number/APSPQQHNCW63DMJPDA24FKZONK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/APSPQQHNCW63DMJPDA24FKZONK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/APSPQQHNCW63DMJPDA24FKZONK/action/storage_attestation","attest_author":"https://pith.science/pith/APSPQQHNCW63DMJPDA24FKZONK/action/author_attestation","sign_citation":"https://pith.science/pith/APSPQQHNCW63DMJPDA24FKZONK/action/citation_signature","submit_replication":"https://pith.science/pith/APSPQQHNCW63DMJPDA24FKZONK/action/replication_record"}},"created_at":"2026-07-05T07:45:41.986238+00:00","updated_at":"2026-07-05T07:45:41.986238+00:00"}