{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:4C24DVGPITIMLEPB5RFWEOCNVM","short_pith_number":"pith:4C24DVGP","schema_version":"1.0","canonical_sha256":"e0b5c1d4cf44d0c591e1ec4b62384dab1ed94755cfffb3369cb65c6122cac774","source":{"kind":"arxiv","id":"2503.02199","version":1},"attestation_state":"computed","paper":{"title":"Words or Vision: Do Vision-Language Models Have Blind Faith in Text?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG","cs.MM"],"primary_cat":"cs.CV","authors_text":"Ailin Deng, Bryan Hooi, Tri Cao, Zhirui Chen","submitted_at":"2025-03-04T02:21:07Z","abstract_excerpt":"Vision-Language Models (VLMs) excel in integrating visual and textual information for vision-centric tasks, but their handling of inconsistencies between modalities is underexplored. We investigate VLMs' modality preferences when faced with visual data and varied textual inputs in vision-centered settings. By introducing textual variations to four vision-centric tasks and evaluating ten Vision-Language Models (VLMs), we discover a \\emph{``blind faith in text''} phenomenon: VLMs disproportionately trust textual data over visual data when inconsistencies arise, leading to significant performance"},"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":"2503.02199","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-03-04T02:21:07Z","cross_cats_sorted":["cs.AI","cs.CL","cs.LG","cs.MM"],"title_canon_sha256":"b8a977c829446e5b7196914e67998c100948dedf07e3155a2ba7f72f725c7a12","abstract_canon_sha256":"39a5e7dcd52e8668e3acb6afac07514e3aa80f6afd535e490e603e6f86584f86"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:23:38.973878Z","signature_b64":"sh6Ezg6R8tuVqtfqdigV7KqnbKKCZ+5H1EM1Nj30qgXvLJsnSuA5XwQySELQRn1IY0SSR2Avl9V5BAeUgNcgBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e0b5c1d4cf44d0c591e1ec4b62384dab1ed94755cfffb3369cb65c6122cac774","last_reissued_at":"2026-07-05T10:23:38.973028Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:23:38.973028Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Words or Vision: Do Vision-Language Models Have Blind Faith in Text?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG","cs.MM"],"primary_cat":"cs.CV","authors_text":"Ailin Deng, Bryan Hooi, Tri Cao, Zhirui Chen","submitted_at":"2025-03-04T02:21:07Z","abstract_excerpt":"Vision-Language Models (VLMs) excel in integrating visual and textual information for vision-centric tasks, but their handling of inconsistencies between modalities is underexplored. We investigate VLMs' modality preferences when faced with visual data and varied textual inputs in vision-centered settings. By introducing textual variations to four vision-centric tasks and evaluating ten Vision-Language Models (VLMs), we discover a \\emph{``blind faith in text''} phenomenon: VLMs disproportionately trust textual data over visual data when inconsistencies arise, leading to significant performance"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.02199","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/2503.02199/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":"2503.02199","created_at":"2026-07-05T10:23:38.973138+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.02199v1","created_at":"2026-07-05T10:23:38.973138+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.02199","created_at":"2026-07-05T10:23:38.973138+00:00"},{"alias_kind":"pith_short_12","alias_value":"4C24DVGPITIM","created_at":"2026-07-05T10:23:38.973138+00:00"},{"alias_kind":"pith_short_16","alias_value":"4C24DVGPITIMLEPB","created_at":"2026-07-05T10:23:38.973138+00:00"},{"alias_kind":"pith_short_8","alias_value":"4C24DVGP","created_at":"2026-07-05T10:23:38.973138+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.08034","citing_title":"Sci-Rho: A Multilingual Visually-Grounded Symbolic Benchmark for STEM Problems","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18803","citing_title":"LLM-as-Judge Framework for Evaluating Tone-Induced Hallucination in Vision-Language Models","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4C24DVGPITIMLEPB5RFWEOCNVM","json":"https://pith.science/pith/4C24DVGPITIMLEPB5RFWEOCNVM.json","graph_json":"https://pith.science/api/pith-number/4C24DVGPITIMLEPB5RFWEOCNVM/graph.json","events_json":"https://pith.science/api/pith-number/4C24DVGPITIMLEPB5RFWEOCNVM/events.json","paper":"https://pith.science/paper/4C24DVGP"},"agent_actions":{"view_html":"https://pith.science/pith/4C24DVGPITIMLEPB5RFWEOCNVM","download_json":"https://pith.science/pith/4C24DVGPITIMLEPB5RFWEOCNVM.json","view_paper":"https://pith.science/paper/4C24DVGP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.02199&json=true","fetch_graph":"https://pith.science/api/pith-number/4C24DVGPITIMLEPB5RFWEOCNVM/graph.json","fetch_events":"https://pith.science/api/pith-number/4C24DVGPITIMLEPB5RFWEOCNVM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4C24DVGPITIMLEPB5RFWEOCNVM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4C24DVGPITIMLEPB5RFWEOCNVM/action/storage_attestation","attest_author":"https://pith.science/pith/4C24DVGPITIMLEPB5RFWEOCNVM/action/author_attestation","sign_citation":"https://pith.science/pith/4C24DVGPITIMLEPB5RFWEOCNVM/action/citation_signature","submit_replication":"https://pith.science/pith/4C24DVGPITIMLEPB5RFWEOCNVM/action/replication_record"}},"created_at":"2026-07-05T10:23:38.973138+00:00","updated_at":"2026-07-05T10:23:38.973138+00:00"}