{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:LT73Z4J3R6ZOFZ3QJFKOQZWJE6","short_pith_number":"pith:LT73Z4J3","schema_version":"1.0","canonical_sha256":"5cffbcf13b8fb2e2e7704954e866c927a118c9e7732c919cb023ec7ceaa61c92","source":{"kind":"arxiv","id":"2401.12425","version":3},"attestation_state":"computed","paper":{"title":"The Neglected Tails in Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Deva Ramanan, James Caverlee, Shubham Parashar, Shu Kong, Tian Liu, Xiangjue Dong, Yanan Li, Zhiqiu Lin","submitted_at":"2024-01-23T01:25:00Z","abstract_excerpt":"Vision-language models (VLMs) excel in zero-shot recognition but their performance varies greatly across different visual concepts. For example, although CLIP achieves impressive accuracy on ImageNet (60-80%), its performance drops below 10% for more than ten concepts like night snake, presumably due to their limited presence in the pretraining data. However, measuring the frequency of concepts in VLMs' large-scale datasets is challenging. We address this by using large language models (LLMs) to count the number of pretraining texts that contain synonyms of these concepts. Our analysis confirm"},"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":"2401.12425","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-01-23T01:25:00Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"57d3cd6a4ab6da240d88f25d3bf2c0bc97ce2007968daa566563cd80149e3fd0","abstract_canon_sha256":"04bf0976de38b14917d9ed60d3ace63a2d50fb55f18d67094d385dd6c506cc9f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:21:48.430534Z","signature_b64":"8Tu0555YQRQo+ZdhzevcFDqNwZeFeiV9app45b+q7FlCU/bCa+Q4a2OSHUUZgV5dfoLzFVBLTlHlnoiLY5DvCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5cffbcf13b8fb2e2e7704954e866c927a118c9e7732c919cb023ec7ceaa61c92","last_reissued_at":"2026-07-05T08:21:48.430009Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:21:48.430009Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Neglected Tails in Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Deva Ramanan, James Caverlee, Shubham Parashar, Shu Kong, Tian Liu, Xiangjue Dong, Yanan Li, Zhiqiu Lin","submitted_at":"2024-01-23T01:25:00Z","abstract_excerpt":"Vision-language models (VLMs) excel in zero-shot recognition but their performance varies greatly across different visual concepts. For example, although CLIP achieves impressive accuracy on ImageNet (60-80%), its performance drops below 10% for more than ten concepts like night snake, presumably due to their limited presence in the pretraining data. However, measuring the frequency of concepts in VLMs' large-scale datasets is challenging. We address this by using large language models (LLMs) to count the number of pretraining texts that contain synonyms of these concepts. Our analysis confirm"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.12425","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/2401.12425/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":"2401.12425","created_at":"2026-07-05T08:21:48.430073+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.12425v3","created_at":"2026-07-05T08:21:48.430073+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.12425","created_at":"2026-07-05T08:21:48.430073+00:00"},{"alias_kind":"pith_short_12","alias_value":"LT73Z4J3R6ZO","created_at":"2026-07-05T08:21:48.430073+00:00"},{"alias_kind":"pith_short_16","alias_value":"LT73Z4J3R6ZOFZ3Q","created_at":"2026-07-05T08:21:48.430073+00:00"},{"alias_kind":"pith_short_8","alias_value":"LT73Z4J3","created_at":"2026-07-05T08:21:48.430073+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.21718","citing_title":"Building a Precise Video Language with Human-AI Oversight","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20366","citing_title":"Mitigating Hallucinations in Large Vision-Language Models without Performance Degradation","ref_index":116,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LT73Z4J3R6ZOFZ3QJFKOQZWJE6","json":"https://pith.science/pith/LT73Z4J3R6ZOFZ3QJFKOQZWJE6.json","graph_json":"https://pith.science/api/pith-number/LT73Z4J3R6ZOFZ3QJFKOQZWJE6/graph.json","events_json":"https://pith.science/api/pith-number/LT73Z4J3R6ZOFZ3QJFKOQZWJE6/events.json","paper":"https://pith.science/paper/LT73Z4J3"},"agent_actions":{"view_html":"https://pith.science/pith/LT73Z4J3R6ZOFZ3QJFKOQZWJE6","download_json":"https://pith.science/pith/LT73Z4J3R6ZOFZ3QJFKOQZWJE6.json","view_paper":"https://pith.science/paper/LT73Z4J3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.12425&json=true","fetch_graph":"https://pith.science/api/pith-number/LT73Z4J3R6ZOFZ3QJFKOQZWJE6/graph.json","fetch_events":"https://pith.science/api/pith-number/LT73Z4J3R6ZOFZ3QJFKOQZWJE6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LT73Z4J3R6ZOFZ3QJFKOQZWJE6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LT73Z4J3R6ZOFZ3QJFKOQZWJE6/action/storage_attestation","attest_author":"https://pith.science/pith/LT73Z4J3R6ZOFZ3QJFKOQZWJE6/action/author_attestation","sign_citation":"https://pith.science/pith/LT73Z4J3R6ZOFZ3QJFKOQZWJE6/action/citation_signature","submit_replication":"https://pith.science/pith/LT73Z4J3R6ZOFZ3QJFKOQZWJE6/action/replication_record"}},"created_at":"2026-07-05T08:21:48.430073+00:00","updated_at":"2026-07-05T08:21:48.430073+00:00"}