{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:MVTJV3JHQLEOFVQAFI5Q3JZF5B","short_pith_number":"pith:MVTJV3JH","schema_version":"1.0","canonical_sha256":"65669aed2782c8e2d6002a3b0da725e850606f71b375d41cd94d39c73fa6930a","source":{"kind":"arxiv","id":"2104.08666","version":2},"attestation_state":"computed","paper":{"title":"Worst of Both Worlds: Biases Compound in Pre-trained Vision-and-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Tejas Srinivasan, Yonatan Bisk","submitted_at":"2021-04-18T00:02:32Z","abstract_excerpt":"Numerous works have analyzed biases in vision and pre-trained language models individually - however, less attention has been paid to how these biases interact in multimodal settings. This work extends text-based bias analysis methods to investigate multimodal language models, and analyzes intra- and inter-modality associations and biases learned by these models. Specifically, we demonstrate that VL-BERT (Su et al., 2020) exhibits gender biases, often preferring to reinforce a stereotype over faithfully describing the visual scene. We demonstrate these findings on a controlled case-study and e"},"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":"2104.08666","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-04-18T00:02:32Z","cross_cats_sorted":[],"title_canon_sha256":"2206e5c15d5f2a405102ecb0b86c5a1266ad686e5826a0e498cdd67508ea3075","abstract_canon_sha256":"0654281d1c90c1a17b8da3d553f514e780437c627f8152328c4c49a3c8f8762c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:24:55.199698Z","signature_b64":"FLGRnSDP3CL057n9xEq/X4h27jREU8/HcJO0fKc9Tkbso3jWb4uVwrwTeDSd7N0EcJl2YXSh0n0AdyM6AkDQBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"65669aed2782c8e2d6002a3b0da725e850606f71b375d41cd94d39c73fa6930a","last_reissued_at":"2026-07-05T04:24:55.199238Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:24:55.199238Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Worst of Both Worlds: Biases Compound in Pre-trained Vision-and-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Tejas Srinivasan, Yonatan Bisk","submitted_at":"2021-04-18T00:02:32Z","abstract_excerpt":"Numerous works have analyzed biases in vision and pre-trained language models individually - however, less attention has been paid to how these biases interact in multimodal settings. This work extends text-based bias analysis methods to investigate multimodal language models, and analyzes intra- and inter-modality associations and biases learned by these models. Specifically, we demonstrate that VL-BERT (Su et al., 2020) exhibits gender biases, often preferring to reinforce a stereotype over faithfully describing the visual scene. We demonstrate these findings on a controlled case-study and e"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2104.08666","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/2104.08666/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":"2104.08666","created_at":"2026-07-05T04:24:55.199295+00:00"},{"alias_kind":"arxiv_version","alias_value":"2104.08666v2","created_at":"2026-07-05T04:24:55.199295+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2104.08666","created_at":"2026-07-05T04:24:55.199295+00:00"},{"alias_kind":"pith_short_12","alias_value":"MVTJV3JHQLEO","created_at":"2026-07-05T04:24:55.199295+00:00"},{"alias_kind":"pith_short_16","alias_value":"MVTJV3JHQLEOFVQA","created_at":"2026-07-05T04:24:55.199295+00:00"},{"alias_kind":"pith_short_8","alias_value":"MVTJV3JH","created_at":"2026-07-05T04:24:55.199295+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2404.18416","citing_title":"Capabilities of Gemini Models in Medicine","ref_index":203,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MVTJV3JHQLEOFVQAFI5Q3JZF5B","json":"https://pith.science/pith/MVTJV3JHQLEOFVQAFI5Q3JZF5B.json","graph_json":"https://pith.science/api/pith-number/MVTJV3JHQLEOFVQAFI5Q3JZF5B/graph.json","events_json":"https://pith.science/api/pith-number/MVTJV3JHQLEOFVQAFI5Q3JZF5B/events.json","paper":"https://pith.science/paper/MVTJV3JH"},"agent_actions":{"view_html":"https://pith.science/pith/MVTJV3JHQLEOFVQAFI5Q3JZF5B","download_json":"https://pith.science/pith/MVTJV3JHQLEOFVQAFI5Q3JZF5B.json","view_paper":"https://pith.science/paper/MVTJV3JH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2104.08666&json=true","fetch_graph":"https://pith.science/api/pith-number/MVTJV3JHQLEOFVQAFI5Q3JZF5B/graph.json","fetch_events":"https://pith.science/api/pith-number/MVTJV3JHQLEOFVQAFI5Q3JZF5B/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MVTJV3JHQLEOFVQAFI5Q3JZF5B/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MVTJV3JHQLEOFVQAFI5Q3JZF5B/action/storage_attestation","attest_author":"https://pith.science/pith/MVTJV3JHQLEOFVQAFI5Q3JZF5B/action/author_attestation","sign_citation":"https://pith.science/pith/MVTJV3JHQLEOFVQAFI5Q3JZF5B/action/citation_signature","submit_replication":"https://pith.science/pith/MVTJV3JHQLEOFVQAFI5Q3JZF5B/action/replication_record"}},"created_at":"2026-07-05T04:24:55.199295+00:00","updated_at":"2026-07-05T04:24:55.199295+00:00"}