{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FRSWWE7GZTHZNB4DZHTOIY5WCS","short_pith_number":"pith:FRSWWE7G","schema_version":"1.0","canonical_sha256":"2c656b13e6cccf968783c9e6e463b614994fb8442b2c3788ed75a73921a3d56c","source":{"kind":"arxiv","id":"2401.02460","version":2},"attestation_state":"computed","paper":{"title":"Improved Zero-Shot Classification by Adapting VLMs with Text Descriptions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Grant Van Horn, Oindrila Saha, Subhransu Maji","submitted_at":"2024-01-04T08:39:13Z","abstract_excerpt":"The zero-shot performance of existing vision-language models (VLMs) such as CLIP is limited by the availability of large-scale, aligned image and text datasets in specific domains. In this work, we leverage two complementary sources of information -- descriptions of categories generated by large language models (LLMs) and abundant, fine-grained image classification datasets -- to improve the zero-shot classification performance of VLMs across fine-grained domains. On the technical side, we develop methods to train VLMs with this \"bag-level\" image-text supervision. We find that simply using 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":"2401.02460","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-01-04T08:39:13Z","cross_cats_sorted":[],"title_canon_sha256":"6a77b13add1698520ca1709ecd3765af59e8e5a7250f3060bade44cb112568e1","abstract_canon_sha256":"ac9a56833ddcfb75aebc82342fee19a9531cd019fae09dbe9e074645e82b33c5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:04:17.305831Z","signature_b64":"JqgKo0UAbJD7ngDtegVp9FI4pb7sD/PgVrzA58KAAJ41QXuEQxoidoOwpnvRdchm4tGHnU5olqKUhVvDXZNDAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2c656b13e6cccf968783c9e6e463b614994fb8442b2c3788ed75a73921a3d56c","last_reissued_at":"2026-07-05T08:04:17.305255Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:04:17.305255Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improved Zero-Shot Classification by Adapting VLMs with Text Descriptions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Grant Van Horn, Oindrila Saha, Subhransu Maji","submitted_at":"2024-01-04T08:39:13Z","abstract_excerpt":"The zero-shot performance of existing vision-language models (VLMs) such as CLIP is limited by the availability of large-scale, aligned image and text datasets in specific domains. In this work, we leverage two complementary sources of information -- descriptions of categories generated by large language models (LLMs) and abundant, fine-grained image classification datasets -- to improve the zero-shot classification performance of VLMs across fine-grained domains. On the technical side, we develop methods to train VLMs with this \"bag-level\" image-text supervision. We find that simply using the"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.02460","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/2401.02460/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":"2401.02460","created_at":"2026-07-05T08:04:17.305351+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.02460v2","created_at":"2026-07-05T08:04:17.305351+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.02460","created_at":"2026-07-05T08:04:17.305351+00:00"},{"alias_kind":"pith_short_12","alias_value":"FRSWWE7GZTHZ","created_at":"2026-07-05T08:04:17.305351+00:00"},{"alias_kind":"pith_short_16","alias_value":"FRSWWE7GZTHZNB4D","created_at":"2026-07-05T08:04:17.305351+00:00"},{"alias_kind":"pith_short_8","alias_value":"FRSWWE7G","created_at":"2026-07-05T08:04:17.305351+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26734","citing_title":"Robust Onion: Peeling Open Vocab Object Detectors Under Noise","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2606.26734","citing_title":"Robust Onion: Peeling Open Vocab Object Detectors Under Noise","ref_index":52,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FRSWWE7GZTHZNB4DZHTOIY5WCS","json":"https://pith.science/pith/FRSWWE7GZTHZNB4DZHTOIY5WCS.json","graph_json":"https://pith.science/api/pith-number/FRSWWE7GZTHZNB4DZHTOIY5WCS/graph.json","events_json":"https://pith.science/api/pith-number/FRSWWE7GZTHZNB4DZHTOIY5WCS/events.json","paper":"https://pith.science/paper/FRSWWE7G"},"agent_actions":{"view_html":"https://pith.science/pith/FRSWWE7GZTHZNB4DZHTOIY5WCS","download_json":"https://pith.science/pith/FRSWWE7GZTHZNB4DZHTOIY5WCS.json","view_paper":"https://pith.science/paper/FRSWWE7G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.02460&json=true","fetch_graph":"https://pith.science/api/pith-number/FRSWWE7GZTHZNB4DZHTOIY5WCS/graph.json","fetch_events":"https://pith.science/api/pith-number/FRSWWE7GZTHZNB4DZHTOIY5WCS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FRSWWE7GZTHZNB4DZHTOIY5WCS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FRSWWE7GZTHZNB4DZHTOIY5WCS/action/storage_attestation","attest_author":"https://pith.science/pith/FRSWWE7GZTHZNB4DZHTOIY5WCS/action/author_attestation","sign_citation":"https://pith.science/pith/FRSWWE7GZTHZNB4DZHTOIY5WCS/action/citation_signature","submit_replication":"https://pith.science/pith/FRSWWE7GZTHZNB4DZHTOIY5WCS/action/replication_record"}},"created_at":"2026-07-05T08:04:17.305351+00:00","updated_at":"2026-07-05T08:04:17.305351+00:00"}