{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:KG5PKRB6AOVIF7JSLJDPWXYCGK","short_pith_number":"pith:KG5PKRB6","schema_version":"1.0","canonical_sha256":"51baf5443e03aa82fd325a46fb5f0232b02e5fdcfc0694b07009bf1e262728da","source":{"kind":"arxiv","id":"2405.18570","version":3},"attestation_state":"computed","paper":{"title":"It's Not a Modality Gap: Characterizing and Addressing the Contrastive Gap","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.IR","cs.LG"],"primary_cat":"cs.CV","authors_text":"Abrar Fahim, Alex Murphy, Alona Fyshe","submitted_at":"2024-05-28T20:28:07Z","abstract_excerpt":"Multi-modal contrastive models such as CLIP achieve state-of-the-art performance in zero-shot classification by embedding input images and texts on a joint representational space. Recently, a modality gap has been reported in two-encoder contrastive models like CLIP, meaning that the image and text embeddings reside in disjoint areas of the latent space. Previous studies suggest that this gap exists due to 1) the cone effect, 2) mismatched pairs in the dataset, and 3) insufficient training. We show that, even when accounting for all these factors, and even when using the same modality, the con"},"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":"2405.18570","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-05-28T20:28:07Z","cross_cats_sorted":["cs.CL","cs.IR","cs.LG"],"title_canon_sha256":"a9294643a214b38b3b38506a4254623bedea964d97bffd35e3ba73449d97a824","abstract_canon_sha256":"84162075770a74229b28eff9012b6a642cff3eb4abd249d5e42da3b378c6bc18"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:28:41.428455Z","signature_b64":"wqwGKiUf+vu6kOsIznMmv9MGblIdYYtuAiehGVdwxYwc73xxL8UqxsXjIIMDzHPyWiC7RSWFKTqb1R5LS8OKDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"51baf5443e03aa82fd325a46fb5f0232b02e5fdcfc0694b07009bf1e262728da","last_reissued_at":"2026-07-05T08:28:41.427981Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:28:41.427981Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"It's Not a Modality Gap: Characterizing and Addressing the Contrastive Gap","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.IR","cs.LG"],"primary_cat":"cs.CV","authors_text":"Abrar Fahim, Alex Murphy, Alona Fyshe","submitted_at":"2024-05-28T20:28:07Z","abstract_excerpt":"Multi-modal contrastive models such as CLIP achieve state-of-the-art performance in zero-shot classification by embedding input images and texts on a joint representational space. Recently, a modality gap has been reported in two-encoder contrastive models like CLIP, meaning that the image and text embeddings reside in disjoint areas of the latent space. Previous studies suggest that this gap exists due to 1) the cone effect, 2) mismatched pairs in the dataset, and 3) insufficient training. We show that, even when accounting for all these factors, and even when using the same modality, the con"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.18570","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/2405.18570/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":"2405.18570","created_at":"2026-07-05T08:28:41.428036+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.18570v3","created_at":"2026-07-05T08:28:41.428036+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.18570","created_at":"2026-07-05T08:28:41.428036+00:00"},{"alias_kind":"pith_short_12","alias_value":"KG5PKRB6AOVI","created_at":"2026-07-05T08:28:41.428036+00:00"},{"alias_kind":"pith_short_16","alias_value":"KG5PKRB6AOVIF7JS","created_at":"2026-07-05T08:28:41.428036+00:00"},{"alias_kind":"pith_short_8","alias_value":"KG5PKRB6","created_at":"2026-07-05T08:28:41.428036+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.29628","citing_title":"COMET: Concept Space Dissection of the Modality Gap in Audio-Text Multimodal Contrastive Embeddings","ref_index":47,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KG5PKRB6AOVIF7JSLJDPWXYCGK","json":"https://pith.science/pith/KG5PKRB6AOVIF7JSLJDPWXYCGK.json","graph_json":"https://pith.science/api/pith-number/KG5PKRB6AOVIF7JSLJDPWXYCGK/graph.json","events_json":"https://pith.science/api/pith-number/KG5PKRB6AOVIF7JSLJDPWXYCGK/events.json","paper":"https://pith.science/paper/KG5PKRB6"},"agent_actions":{"view_html":"https://pith.science/pith/KG5PKRB6AOVIF7JSLJDPWXYCGK","download_json":"https://pith.science/pith/KG5PKRB6AOVIF7JSLJDPWXYCGK.json","view_paper":"https://pith.science/paper/KG5PKRB6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.18570&json=true","fetch_graph":"https://pith.science/api/pith-number/KG5PKRB6AOVIF7JSLJDPWXYCGK/graph.json","fetch_events":"https://pith.science/api/pith-number/KG5PKRB6AOVIF7JSLJDPWXYCGK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KG5PKRB6AOVIF7JSLJDPWXYCGK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KG5PKRB6AOVIF7JSLJDPWXYCGK/action/storage_attestation","attest_author":"https://pith.science/pith/KG5PKRB6AOVIF7JSLJDPWXYCGK/action/author_attestation","sign_citation":"https://pith.science/pith/KG5PKRB6AOVIF7JSLJDPWXYCGK/action/citation_signature","submit_replication":"https://pith.science/pith/KG5PKRB6AOVIF7JSLJDPWXYCGK/action/replication_record"}},"created_at":"2026-07-05T08:28:41.428036+00:00","updated_at":"2026-07-05T08:28:41.428036+00:00"}