{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:MY4CORRQU4QJCQL4SBLSJMBBIB","short_pith_number":"pith:MY4CORRQ","schema_version":"1.0","canonical_sha256":"6638274630a72091417c905724b021404474b35de989ba203be098ca1f5c2e07","source":{"kind":"arxiv","id":"2310.09562","version":2},"attestation_state":"computed","paper":{"title":"Does CLIP's Generalization Performance Mainly Stem from High Train-Test Similarity?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Evgenia Rusak, Matthias Bethge, Prasanna Mayilvahanan, Thadd\\\"aus Wiedemer, Wieland Brendel","submitted_at":"2023-10-14T11:24:28Z","abstract_excerpt":"Foundation models like CLIP are trained on hundreds of millions of samples and effortlessly generalize to new tasks and inputs. Out of the box, CLIP shows stellar zero-shot and few-shot capabilities on a wide range of out-of-distribution (OOD) benchmarks, which prior works attribute mainly to today's large and comprehensive training dataset (like LAION). However, it is questionable how meaningful terms like out-of-distribution generalization are for CLIP as it seems likely that web-scale datasets like LAION simply contain many samples that are similar to common OOD benchmarks originally design"},"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":"2310.09562","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2023-10-14T11:24:28Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"a24b86d0b65ded55e369958656bcfb6ed8cb606626a21f53269a4d557ab2b654","abstract_canon_sha256":"656ab2e907aeeccb090586f81e9f5284cdafa88390f46468ad3f201bc3b94739"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:56:17.825618Z","signature_b64":"RO0c7Czw/ZEqZcJ869nnNRLOIrxfAorcCOwaQHpFiAdww1WzmFswmOarhLCBuiideKux0NFdRTM2S+7KWZh4Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6638274630a72091417c905724b021404474b35de989ba203be098ca1f5c2e07","last_reissued_at":"2026-07-05T07:56:17.825107Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:56:17.825107Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Does CLIP's Generalization Performance Mainly Stem from High Train-Test Similarity?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Evgenia Rusak, Matthias Bethge, Prasanna Mayilvahanan, Thadd\\\"aus Wiedemer, Wieland Brendel","submitted_at":"2023-10-14T11:24:28Z","abstract_excerpt":"Foundation models like CLIP are trained on hundreds of millions of samples and effortlessly generalize to new tasks and inputs. Out of the box, CLIP shows stellar zero-shot and few-shot capabilities on a wide range of out-of-distribution (OOD) benchmarks, which prior works attribute mainly to today's large and comprehensive training dataset (like LAION). However, it is questionable how meaningful terms like out-of-distribution generalization are for CLIP as it seems likely that web-scale datasets like LAION simply contain many samples that are similar to common OOD benchmarks originally design"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.09562","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/2310.09562/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":"2310.09562","created_at":"2026-07-05T07:56:17.825166+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.09562v2","created_at":"2026-07-05T07:56:17.825166+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.09562","created_at":"2026-07-05T07:56:17.825166+00:00"},{"alias_kind":"pith_short_12","alias_value":"MY4CORRQU4QJ","created_at":"2026-07-05T07:56:17.825166+00:00"},{"alias_kind":"pith_short_16","alias_value":"MY4CORRQU4QJCQL4","created_at":"2026-07-05T07:56:17.825166+00:00"},{"alias_kind":"pith_short_8","alias_value":"MY4CORRQ","created_at":"2026-07-05T07:56:17.825166+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2502.12120","citing_title":"LLMs on the Line: Data Determines Loss-to-Loss Scaling Laws","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2506.16950","citing_title":"LAION-C: An Out-of-Distribution Benchmark for Web-Scale Vision Models","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2509.21173","citing_title":"Less Precise Can Be More Reliable: A Systematic Evaluation of Quantization's Impact on VLMs Beyond Accuracy","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2509.21173","citing_title":"Less Precise Can Be More Reliable: A Systematic Evaluation of Quantization's Impact on VLMs Beyond Accuracy","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MY4CORRQU4QJCQL4SBLSJMBBIB","json":"https://pith.science/pith/MY4CORRQU4QJCQL4SBLSJMBBIB.json","graph_json":"https://pith.science/api/pith-number/MY4CORRQU4QJCQL4SBLSJMBBIB/graph.json","events_json":"https://pith.science/api/pith-number/MY4CORRQU4QJCQL4SBLSJMBBIB/events.json","paper":"https://pith.science/paper/MY4CORRQ"},"agent_actions":{"view_html":"https://pith.science/pith/MY4CORRQU4QJCQL4SBLSJMBBIB","download_json":"https://pith.science/pith/MY4CORRQU4QJCQL4SBLSJMBBIB.json","view_paper":"https://pith.science/paper/MY4CORRQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.09562&json=true","fetch_graph":"https://pith.science/api/pith-number/MY4CORRQU4QJCQL4SBLSJMBBIB/graph.json","fetch_events":"https://pith.science/api/pith-number/MY4CORRQU4QJCQL4SBLSJMBBIB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MY4CORRQU4QJCQL4SBLSJMBBIB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MY4CORRQU4QJCQL4SBLSJMBBIB/action/storage_attestation","attest_author":"https://pith.science/pith/MY4CORRQU4QJCQL4SBLSJMBBIB/action/author_attestation","sign_citation":"https://pith.science/pith/MY4CORRQU4QJCQL4SBLSJMBBIB/action/citation_signature","submit_replication":"https://pith.science/pith/MY4CORRQU4QJCQL4SBLSJMBBIB/action/replication_record"}},"created_at":"2026-07-05T07:56:17.825166+00:00","updated_at":"2026-07-05T07:56:17.825166+00:00"}