{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:WAJXJGYDIA6A7TBJCY7ZHOJOM2","short_pith_number":"pith:WAJXJGYD","schema_version":"1.0","canonical_sha256":"b013749b03403c0fcc29163f93b92e668c63cba71b00483e2f151ba469db1458","source":{"kind":"arxiv","id":"2406.09188","version":2},"attestation_state":"computed","paper":{"title":"An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.IR"],"primary_cat":"cs.CV","authors_text":"Jaeseok Byun, Sanghyuk Chun, Seokhyeon Jeong, Taesup Moon, Wonjae Kim","submitted_at":"2024-06-13T14:49:28Z","abstract_excerpt":"Composed Image Retrieval (CIR) aims to retrieve a target image based on a reference image and conditioning text, enabling controllable image searches. The mainstream Zero-Shot (ZS) CIR methods bypass the need for expensive training CIR triplets by projecting image embeddings into the text token embedding space, forming a composed query for retrieval. However, we highlight an inherent limitation in these projection-based CIR: a task discrepancy of text encoders between the original pre-training task of the encoders (text $\\leftrightarrow$ image) and the target CIR task (image + text $\\leftright"},"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":"2406.09188","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2024-06-13T14:49:28Z","cross_cats_sorted":["cs.IR"],"title_canon_sha256":"6a0966add29e62c443663bf366c626ae3c4e50209156d01f0c867f7caa580997","abstract_canon_sha256":"d91a4f969623efb608c4647b587f5fb9357a302ae833c33f007a4094c0370331"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:33:04.042496Z","signature_b64":"C+veSUc4aG5ZzfY9vPXKCh9Cvmp0tOgi50mGa+BW4fUyVPmsto1f6mwD5jUrtS5ViXP23Or1OymEbOtSQTz+DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b013749b03403c0fcc29163f93b92e668c63cba71b00483e2f151ba469db1458","last_reissued_at":"2026-07-05T10:33:04.041619Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:33:04.041619Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.IR"],"primary_cat":"cs.CV","authors_text":"Jaeseok Byun, Sanghyuk Chun, Seokhyeon Jeong, Taesup Moon, Wonjae Kim","submitted_at":"2024-06-13T14:49:28Z","abstract_excerpt":"Composed Image Retrieval (CIR) aims to retrieve a target image based on a reference image and conditioning text, enabling controllable image searches. The mainstream Zero-Shot (ZS) CIR methods bypass the need for expensive training CIR triplets by projecting image embeddings into the text token embedding space, forming a composed query for retrieval. However, we highlight an inherent limitation in these projection-based CIR: a task discrepancy of text encoders between the original pre-training task of the encoders (text $\\leftrightarrow$ image) and the target CIR task (image + text $\\leftright"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.09188","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/2406.09188/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":"2406.09188","created_at":"2026-07-05T10:33:04.041722+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.09188v2","created_at":"2026-07-05T10:33:04.041722+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.09188","created_at":"2026-07-05T10:33:04.041722+00:00"},{"alias_kind":"pith_short_12","alias_value":"WAJXJGYDIA6A","created_at":"2026-07-05T10:33:04.041722+00:00"},{"alias_kind":"pith_short_16","alias_value":"WAJXJGYDIA6A7TBJ","created_at":"2026-07-05T10:33:04.041722+00:00"},{"alias_kind":"pith_short_8","alias_value":"WAJXJGYD","created_at":"2026-07-05T10:33:04.041722+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2412.05756","citing_title":"Compositional Image Retrieval via Instruction-Aware Contrastive Learning","ref_index":9,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WAJXJGYDIA6A7TBJCY7ZHOJOM2","json":"https://pith.science/pith/WAJXJGYDIA6A7TBJCY7ZHOJOM2.json","graph_json":"https://pith.science/api/pith-number/WAJXJGYDIA6A7TBJCY7ZHOJOM2/graph.json","events_json":"https://pith.science/api/pith-number/WAJXJGYDIA6A7TBJCY7ZHOJOM2/events.json","paper":"https://pith.science/paper/WAJXJGYD"},"agent_actions":{"view_html":"https://pith.science/pith/WAJXJGYDIA6A7TBJCY7ZHOJOM2","download_json":"https://pith.science/pith/WAJXJGYDIA6A7TBJCY7ZHOJOM2.json","view_paper":"https://pith.science/paper/WAJXJGYD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.09188&json=true","fetch_graph":"https://pith.science/api/pith-number/WAJXJGYDIA6A7TBJCY7ZHOJOM2/graph.json","fetch_events":"https://pith.science/api/pith-number/WAJXJGYDIA6A7TBJCY7ZHOJOM2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WAJXJGYDIA6A7TBJCY7ZHOJOM2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WAJXJGYDIA6A7TBJCY7ZHOJOM2/action/storage_attestation","attest_author":"https://pith.science/pith/WAJXJGYDIA6A7TBJCY7ZHOJOM2/action/author_attestation","sign_citation":"https://pith.science/pith/WAJXJGYDIA6A7TBJCY7ZHOJOM2/action/citation_signature","submit_replication":"https://pith.science/pith/WAJXJGYDIA6A7TBJCY7ZHOJOM2/action/replication_record"}},"created_at":"2026-07-05T10:33:04.041722+00:00","updated_at":"2026-07-05T10:33:04.041722+00:00"}