{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:V7B474SP3EI6VV7XGXRSWH7XK4","short_pith_number":"pith:V7B474SP","schema_version":"1.0","canonical_sha256":"afc3cff24fd911ead7f735e32b1ff75709329aec7847135cde4b0de68f216e31","source":{"kind":"arxiv","id":"2207.07635","version":1},"attestation_state":"computed","paper":{"title":"Is a Caption Worth a Thousand Images? A Controlled Study for Representation Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","stat.ML"],"primary_cat":"cs.CV","authors_text":"Percy Liang, Rohan Taori, Shibani Santurkar, Tatsunori Hashimoto, Yann Dubois","submitted_at":"2022-07-15T17:50:51Z","abstract_excerpt":"The development of CLIP [Radford et al., 2021] has sparked a debate on whether language supervision can result in vision models with more transferable representations than traditional image-only methods. Our work studies this question through a carefully controlled comparison of two approaches in terms of their ability to learn representations that generalize to downstream classification tasks. We find that when the pre-training dataset meets certain criteria -- it is sufficiently large and contains descriptive captions with low variability -- image-only methods do not match CLIP's transfer pe"},"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":"2207.07635","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2022-07-15T17:50:51Z","cross_cats_sorted":["cs.LG","stat.ML"],"title_canon_sha256":"6ec100db6356550eb6d0ab6d1ed772beb9ee02e62a04e4d05d6430668606e2fa","abstract_canon_sha256":"0b0305dded64f5e636f7d30ff7d30ed0e5cf3dd636383c232cc6420695631395"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:40:34.699592Z","signature_b64":"YlPaOxD6iLgQfVj5LyS4EY3ulPKgwAeZwV0/9Swe7AtlvQqtjwmQB++oZWdRW3udYUHYgmkvmX6G+b6g/kqVBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"afc3cff24fd911ead7f735e32b1ff75709329aec7847135cde4b0de68f216e31","last_reissued_at":"2026-07-05T04:40:34.699229Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:40:34.699229Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Is a Caption Worth a Thousand Images? A Controlled Study for Representation Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG","stat.ML"],"primary_cat":"cs.CV","authors_text":"Percy Liang, Rohan Taori, Shibani Santurkar, Tatsunori Hashimoto, Yann Dubois","submitted_at":"2022-07-15T17:50:51Z","abstract_excerpt":"The development of CLIP [Radford et al., 2021] has sparked a debate on whether language supervision can result in vision models with more transferable representations than traditional image-only methods. Our work studies this question through a carefully controlled comparison of two approaches in terms of their ability to learn representations that generalize to downstream classification tasks. We find that when the pre-training dataset meets certain criteria -- it is sufficiently large and contains descriptive captions with low variability -- image-only methods do not match CLIP's transfer pe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2207.07635","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/2207.07635/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":"2207.07635","created_at":"2026-07-05T04:40:34.699287+00:00"},{"alias_kind":"arxiv_version","alias_value":"2207.07635v1","created_at":"2026-07-05T04:40:34.699287+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2207.07635","created_at":"2026-07-05T04:40:34.699287+00:00"},{"alias_kind":"pith_short_12","alias_value":"V7B474SP3EI6","created_at":"2026-07-05T04:40:34.699287+00:00"},{"alias_kind":"pith_short_16","alias_value":"V7B474SP3EI6VV7X","created_at":"2026-07-05T04:40:34.699287+00:00"},{"alias_kind":"pith_short_8","alias_value":"V7B474SP","created_at":"2026-07-05T04:40:34.699287+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2309.16671","citing_title":"Demystifying CLIP Data","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2308.01390","citing_title":"OpenFlamingo: An Open-Source Framework for Training Large Autoregressive Vision-Language Models","ref_index":31,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/V7B474SP3EI6VV7XGXRSWH7XK4","json":"https://pith.science/pith/V7B474SP3EI6VV7XGXRSWH7XK4.json","graph_json":"https://pith.science/api/pith-number/V7B474SP3EI6VV7XGXRSWH7XK4/graph.json","events_json":"https://pith.science/api/pith-number/V7B474SP3EI6VV7XGXRSWH7XK4/events.json","paper":"https://pith.science/paper/V7B474SP"},"agent_actions":{"view_html":"https://pith.science/pith/V7B474SP3EI6VV7XGXRSWH7XK4","download_json":"https://pith.science/pith/V7B474SP3EI6VV7XGXRSWH7XK4.json","view_paper":"https://pith.science/paper/V7B474SP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2207.07635&json=true","fetch_graph":"https://pith.science/api/pith-number/V7B474SP3EI6VV7XGXRSWH7XK4/graph.json","fetch_events":"https://pith.science/api/pith-number/V7B474SP3EI6VV7XGXRSWH7XK4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V7B474SP3EI6VV7XGXRSWH7XK4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V7B474SP3EI6VV7XGXRSWH7XK4/action/storage_attestation","attest_author":"https://pith.science/pith/V7B474SP3EI6VV7XGXRSWH7XK4/action/author_attestation","sign_citation":"https://pith.science/pith/V7B474SP3EI6VV7XGXRSWH7XK4/action/citation_signature","submit_replication":"https://pith.science/pith/V7B474SP3EI6VV7XGXRSWH7XK4/action/replication_record"}},"created_at":"2026-07-05T04:40:34.699287+00:00","updated_at":"2026-07-05T04:40:34.699287+00:00"}