{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:3EH2NPPAYZ2CZPCPBKOKOHA34E","short_pith_number":"pith:3EH2NPPA","schema_version":"1.0","canonical_sha256":"d90fa6bde0c6742cbc4f0a9ca71c1be12a6042a8b9449c0b028ed50cf6ef6405","source":{"kind":"arxiv","id":"2304.03659","version":3},"attestation_state":"computed","paper":{"title":"Probing Conceptual Understanding of Large Visual-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ajay Divakaran, Jared Claypoole, Madeline Schiappa, Michael Cogswell, Raiyaan Abdullah, Shehreen Azad, Yogesh Rawat","submitted_at":"2023-04-07T14:26:11Z","abstract_excerpt":"In recent years large visual-language (V+L) models have achieved great success in various downstream tasks. However, it is not well studied whether these models have a conceptual grasp of the visual content. In this work we focus on conceptual understanding of these large V+L models. To facilitate this study, we propose novel benchmarking datasets for probing three different aspects of content understanding, 1) \\textit{relations}, 2) \\textit{composition}, and 3) \\textit{context}. Our probes are grounded in cognitive science and help determine if a V+L model can, for example, determine if snow "},"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":"2304.03659","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2023-04-07T14:26:11Z","cross_cats_sorted":[],"title_canon_sha256":"4e0cad2155293d0ffe3a14615d7966187212538da3d80e628cb79fc5140c102d","abstract_canon_sha256":"e27e275e1025be98b1c1568a637ff681df85bb1d2fa36a109b40f09adede4fc5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:12:25.635127Z","signature_b64":"FvkKkAdXdwFGt8DijlcNlU621gIe2sTsQmjcsd8WVdTKwXZ1hCjWL8HAG0QEqRbjK3EPQC8RJHtpR7/9h+UkAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d90fa6bde0c6742cbc4f0a9ca71c1be12a6042a8b9449c0b028ed50cf6ef6405","last_reissued_at":"2026-07-05T08:12:25.634703Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:12:25.634703Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Probing Conceptual Understanding of Large Visual-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Ajay Divakaran, Jared Claypoole, Madeline Schiappa, Michael Cogswell, Raiyaan Abdullah, Shehreen Azad, Yogesh Rawat","submitted_at":"2023-04-07T14:26:11Z","abstract_excerpt":"In recent years large visual-language (V+L) models have achieved great success in various downstream tasks. However, it is not well studied whether these models have a conceptual grasp of the visual content. In this work we focus on conceptual understanding of these large V+L models. To facilitate this study, we propose novel benchmarking datasets for probing three different aspects of content understanding, 1) \\textit{relations}, 2) \\textit{composition}, and 3) \\textit{context}. Our probes are grounded in cognitive science and help determine if a V+L model can, for example, determine if snow "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2304.03659","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/2304.03659/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":"2304.03659","created_at":"2026-07-05T08:12:25.634761+00:00"},{"alias_kind":"arxiv_version","alias_value":"2304.03659v3","created_at":"2026-07-05T08:12:25.634761+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2304.03659","created_at":"2026-07-05T08:12:25.634761+00:00"},{"alias_kind":"pith_short_12","alias_value":"3EH2NPPAYZ2C","created_at":"2026-07-05T08:12:25.634761+00:00"},{"alias_kind":"pith_short_16","alias_value":"3EH2NPPAYZ2CZPCP","created_at":"2026-07-05T08:12:25.634761+00:00"},{"alias_kind":"pith_short_8","alias_value":"3EH2NPPA","created_at":"2026-07-05T08:12:25.634761+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.03950","citing_title":"LR0.FM: Low-Res Benchmark and Improving Robustness for Zero-Shot Classification in Foundation Models","ref_index":60,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3EH2NPPAYZ2CZPCPBKOKOHA34E","json":"https://pith.science/pith/3EH2NPPAYZ2CZPCPBKOKOHA34E.json","graph_json":"https://pith.science/api/pith-number/3EH2NPPAYZ2CZPCPBKOKOHA34E/graph.json","events_json":"https://pith.science/api/pith-number/3EH2NPPAYZ2CZPCPBKOKOHA34E/events.json","paper":"https://pith.science/paper/3EH2NPPA"},"agent_actions":{"view_html":"https://pith.science/pith/3EH2NPPAYZ2CZPCPBKOKOHA34E","download_json":"https://pith.science/pith/3EH2NPPAYZ2CZPCPBKOKOHA34E.json","view_paper":"https://pith.science/paper/3EH2NPPA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2304.03659&json=true","fetch_graph":"https://pith.science/api/pith-number/3EH2NPPAYZ2CZPCPBKOKOHA34E/graph.json","fetch_events":"https://pith.science/api/pith-number/3EH2NPPAYZ2CZPCPBKOKOHA34E/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3EH2NPPAYZ2CZPCPBKOKOHA34E/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3EH2NPPAYZ2CZPCPBKOKOHA34E/action/storage_attestation","attest_author":"https://pith.science/pith/3EH2NPPAYZ2CZPCPBKOKOHA34E/action/author_attestation","sign_citation":"https://pith.science/pith/3EH2NPPAYZ2CZPCPBKOKOHA34E/action/citation_signature","submit_replication":"https://pith.science/pith/3EH2NPPAYZ2CZPCPBKOKOHA34E/action/replication_record"}},"created_at":"2026-07-05T08:12:25.634761+00:00","updated_at":"2026-07-05T08:12:25.634761+00:00"}