{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:B7LQEOVDOPO4O57IUUWRGMDN52","short_pith_number":"pith:B7LQEOVD","schema_version":"1.0","canonical_sha256":"0fd7023aa373ddc777e8a52d13306deea962ab27a345368c18b7695c1ac5db92","source":{"kind":"arxiv","id":"2203.13457","version":2},"attestation_state":"computed","paper":{"title":"Chaos is a Ladder: A New Theoretical Understanding of Contrastive Learning via Augmentation Overlap","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Jiansheng Yang, Qi Zhang, Yifei Wang, Yisen Wang, Zhouchen Lin","submitted_at":"2022-03-25T05:36:26Z","abstract_excerpt":"Recently, contrastive learning has risen to be a promising approach for large-scale self-supervised learning. However, theoretical understanding of how it works is still unclear. In this paper, we propose a new guarantee on the downstream performance without resorting to the conditional independence assumption that is widely adopted in previous work but hardly holds in practice. Our new theory hinges on the insight that the support of different intra-class samples will become more overlapped under aggressive data augmentations, thus simply aligning the positive samples (augmented views of the "},"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":"2203.13457","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2022-03-25T05:36:26Z","cross_cats_sorted":["cs.CV","stat.ML"],"title_canon_sha256":"5e93ca7fbefe4e744a3989fa0cc16f0eb2bc9f9a0a6d3775943996e78bad69f0","abstract_canon_sha256":"7d3cd14c576b81eb666dc47299905db71b28310f713f2bc25d1276528749bb12"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:26:48.815481Z","signature_b64":"ZAeInG5ZHBw6LVouerY0B8nKgpFiAACEplCYccoLI2C4Jb7U0J1fwY+WvXqtxHwt+idH74PiUMtnsMEt6Az+DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0fd7023aa373ddc777e8a52d13306deea962ab27a345368c18b7695c1ac5db92","last_reissued_at":"2026-07-05T04:26:48.814947Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:26:48.814947Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Chaos is a Ladder: A New Theoretical Understanding of Contrastive Learning via Augmentation Overlap","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV","stat.ML"],"primary_cat":"cs.LG","authors_text":"Jiansheng Yang, Qi Zhang, Yifei Wang, Yisen Wang, Zhouchen Lin","submitted_at":"2022-03-25T05:36:26Z","abstract_excerpt":"Recently, contrastive learning has risen to be a promising approach for large-scale self-supervised learning. However, theoretical understanding of how it works is still unclear. In this paper, we propose a new guarantee on the downstream performance without resorting to the conditional independence assumption that is widely adopted in previous work but hardly holds in practice. Our new theory hinges on the insight that the support of different intra-class samples will become more overlapped under aggressive data augmentations, thus simply aligning the positive samples (augmented views of the "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2203.13457","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/2203.13457/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":"2203.13457","created_at":"2026-07-05T04:26:48.815007+00:00"},{"alias_kind":"arxiv_version","alias_value":"2203.13457v2","created_at":"2026-07-05T04:26:48.815007+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2203.13457","created_at":"2026-07-05T04:26:48.815007+00:00"},{"alias_kind":"pith_short_12","alias_value":"B7LQEOVDOPO4","created_at":"2026-07-05T04:26:48.815007+00:00"},{"alias_kind":"pith_short_16","alias_value":"B7LQEOVDOPO4O57I","created_at":"2026-07-05T04:26:48.815007+00:00"},{"alias_kind":"pith_short_8","alias_value":"B7LQEOVD","created_at":"2026-07-05T04:26:48.815007+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.22873","citing_title":"SingGuard: A Policy-Adaptive Multimodal LLM Guardrail with Dynamic Reasoning","ref_index":110,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01821","citing_title":"Pattern-Calibrated Multimodal Prediction under Blockwise Missingness","ref_index":135,"is_internal_anchor":false},{"citing_arxiv_id":"2606.22873","citing_title":"SingGuard: A Policy-Adaptive Multimodal LLM Guardrail with Dynamic Reasoning","ref_index":110,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/B7LQEOVDOPO4O57IUUWRGMDN52","json":"https://pith.science/pith/B7LQEOVDOPO4O57IUUWRGMDN52.json","graph_json":"https://pith.science/api/pith-number/B7LQEOVDOPO4O57IUUWRGMDN52/graph.json","events_json":"https://pith.science/api/pith-number/B7LQEOVDOPO4O57IUUWRGMDN52/events.json","paper":"https://pith.science/paper/B7LQEOVD"},"agent_actions":{"view_html":"https://pith.science/pith/B7LQEOVDOPO4O57IUUWRGMDN52","download_json":"https://pith.science/pith/B7LQEOVDOPO4O57IUUWRGMDN52.json","view_paper":"https://pith.science/paper/B7LQEOVD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2203.13457&json=true","fetch_graph":"https://pith.science/api/pith-number/B7LQEOVDOPO4O57IUUWRGMDN52/graph.json","fetch_events":"https://pith.science/api/pith-number/B7LQEOVDOPO4O57IUUWRGMDN52/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/B7LQEOVDOPO4O57IUUWRGMDN52/action/timestamp_anchor","attest_storage":"https://pith.science/pith/B7LQEOVDOPO4O57IUUWRGMDN52/action/storage_attestation","attest_author":"https://pith.science/pith/B7LQEOVDOPO4O57IUUWRGMDN52/action/author_attestation","sign_citation":"https://pith.science/pith/B7LQEOVDOPO4O57IUUWRGMDN52/action/citation_signature","submit_replication":"https://pith.science/pith/B7LQEOVDOPO4O57IUUWRGMDN52/action/replication_record"}},"created_at":"2026-07-05T04:26:48.815007+00:00","updated_at":"2026-07-05T04:26:48.815007+00:00"}