{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:3NO6CUIOPM7TJZYXWGVJOL3WNS","short_pith_number":"pith:3NO6CUIO","schema_version":"1.0","canonical_sha256":"db5de1510e7b3f34e717b1aa972f766cbceff24bb01abecaead6357aec6fbb4a","source":{"kind":"arxiv","id":"2401.08567","version":1},"attestation_state":"computed","paper":{"title":"Connect, Collapse, Corrupt: Learning Cross-Modal Tasks with Uni-Modal Data","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Elaine Sui, Serena Yeung-Levy, Yuhui Zhang","submitted_at":"2024-01-16T18:52:27Z","abstract_excerpt":"Building cross-modal applications is challenging due to limited paired multi-modal data. Recent works have shown that leveraging a pre-trained multi-modal contrastive representation space enables cross-modal tasks to be learned from uni-modal data. This is based on the assumption that contrastive optimization makes embeddings from different modalities interchangeable. However, this assumption is under-explored due to the poorly understood geometry of the multi-modal contrastive space, where a modality gap exists. In our study, we provide a theoretical explanation of this space's geometry and i"},"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":"2401.08567","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-01-16T18:52:27Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CV"],"title_canon_sha256":"09b6d37bd5de353c50ec06e18e2dfbf1be07d947c92c0ea68b81d977b08e5dd7","abstract_canon_sha256":"3f7c911837a35e319e60525a302571597e0b44b24c317d4f9441d51fd6489347"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:34:13.505483Z","signature_b64":"Vi2vJdyrpyO5rl99HqASVxhv42P5+jZ8QGRHmxKQtM7wlokGyUpAjUPEg1TnS0PRu0xQyh3L2MKEjb85WqTuAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"db5de1510e7b3f34e717b1aa972f766cbceff24bb01abecaead6357aec6fbb4a","last_reissued_at":"2026-07-05T07:34:13.505052Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:34:13.505052Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Connect, Collapse, Corrupt: Learning Cross-Modal Tasks with Uni-Modal Data","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Elaine Sui, Serena Yeung-Levy, Yuhui Zhang","submitted_at":"2024-01-16T18:52:27Z","abstract_excerpt":"Building cross-modal applications is challenging due to limited paired multi-modal data. Recent works have shown that leveraging a pre-trained multi-modal contrastive representation space enables cross-modal tasks to be learned from uni-modal data. This is based on the assumption that contrastive optimization makes embeddings from different modalities interchangeable. However, this assumption is under-explored due to the poorly understood geometry of the multi-modal contrastive space, where a modality gap exists. In our study, we provide a theoretical explanation of this space's geometry and i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.08567","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/2401.08567/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":"2401.08567","created_at":"2026-07-05T07:34:13.505108+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.08567v1","created_at":"2026-07-05T07:34:13.505108+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.08567","created_at":"2026-07-05T07:34:13.505108+00:00"},{"alias_kind":"pith_short_12","alias_value":"3NO6CUIOPM7T","created_at":"2026-07-05T07:34:13.505108+00:00"},{"alias_kind":"pith_short_16","alias_value":"3NO6CUIOPM7TJZYX","created_at":"2026-07-05T07:34:13.505108+00:00"},{"alias_kind":"pith_short_8","alias_value":"3NO6CUIO","created_at":"2026-07-05T07:34:13.505108+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2602.07026","citing_title":"Modality Gap-Driven Subspace Alignment Training Paradigm For Multimodal Large Language Models","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07825","citing_title":"Anisotropic Modality Align","ref_index":21,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3NO6CUIOPM7TJZYXWGVJOL3WNS","json":"https://pith.science/pith/3NO6CUIOPM7TJZYXWGVJOL3WNS.json","graph_json":"https://pith.science/api/pith-number/3NO6CUIOPM7TJZYXWGVJOL3WNS/graph.json","events_json":"https://pith.science/api/pith-number/3NO6CUIOPM7TJZYXWGVJOL3WNS/events.json","paper":"https://pith.science/paper/3NO6CUIO"},"agent_actions":{"view_html":"https://pith.science/pith/3NO6CUIOPM7TJZYXWGVJOL3WNS","download_json":"https://pith.science/pith/3NO6CUIOPM7TJZYXWGVJOL3WNS.json","view_paper":"https://pith.science/paper/3NO6CUIO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.08567&json=true","fetch_graph":"https://pith.science/api/pith-number/3NO6CUIOPM7TJZYXWGVJOL3WNS/graph.json","fetch_events":"https://pith.science/api/pith-number/3NO6CUIOPM7TJZYXWGVJOL3WNS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3NO6CUIOPM7TJZYXWGVJOL3WNS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3NO6CUIOPM7TJZYXWGVJOL3WNS/action/storage_attestation","attest_author":"https://pith.science/pith/3NO6CUIOPM7TJZYXWGVJOL3WNS/action/author_attestation","sign_citation":"https://pith.science/pith/3NO6CUIOPM7TJZYXWGVJOL3WNS/action/citation_signature","submit_replication":"https://pith.science/pith/3NO6CUIOPM7TJZYXWGVJOL3WNS/action/replication_record"}},"created_at":"2026-07-05T07:34:13.505108+00:00","updated_at":"2026-07-05T07:34:13.505108+00:00"}