{"paper":{"title":"Pix2Fact: When Vision Is Not Enough -- Benchmarking Fine-Grained VQA with Web Verification on High-Resolution Real-World Scenes","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"Current top vision-language models reach only 51.7 percent accuracy on questions that demand both precise visual details from high-resolution scenes and external knowledge verification.","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Bingzhang Wang, Bofei Zhang, Cong Zhang, Qiaofeng Zheng, Yew-Soon Ong, Yifan Jiang, Yifan Yang","submitted_at":"2026-01-31T08:18:34Z","abstract_excerpt":"Despite progress on general tasks, vision-language models (VLMs) still struggle with challenges that demand both fine-grained visual grounding and external knowledge, a synergy overlooked by existing benchmarks that evaluate these abilities in isolation. To fill this void, we introduce Pix2Fact, a visual question-answering benchmark designed to assess expert-level visual perception and knowledge search. Pix2Fact comprises 1,000 high-resolution (4K+) images spanning eight scenarios. Its questions and answers are meticulously crafted by PhD-holding annotators from top global universities across "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"the most advanced model (Gemini-3.1-Pro) achieves only 51.7% average accuracy, even with access to visual ground truth and search tools.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The questions and answers produced by PhD annotators from top universities faithfully represent expert-level challenges of fine-grained visual grounding plus external knowledge without introducing systematic biases or inconsistent difficulty.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Pix2Fact benchmark shows top VLMs achieve only 51.7 percent accuracy on fine-grained visual questions needing both detailed image grounding and web-verified external knowledge.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Current top vision-language models reach only 51.7 percent accuracy on questions that demand both precise visual details from high-resolution scenes and external knowledge verification.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"216b1ca2821a3dd597ac82d45526274301a1869b4601e9ea8532bb2c4b01b06b"},"source":{"id":"2602.00593","kind":"arxiv","version":3},"verdict":{"id":"18d063c2-dd20-497c-8c2a-5621250bc686","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T09:11:44.471030Z","strongest_claim":"the most advanced model (Gemini-3.1-Pro) achieves only 51.7% average accuracy, even with access to visual ground truth and search tools.","one_line_summary":"Pix2Fact benchmark shows top VLMs achieve only 51.7 percent accuracy on fine-grained visual questions needing both detailed image grounding and web-verified external knowledge.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The questions and answers produced by PhD annotators from top universities faithfully represent expert-level challenges of fine-grained visual grounding plus external knowledge without introducing systematic biases or inconsistent difficulty.","pith_extraction_headline":"Current top vision-language models reach only 51.7 percent accuracy on questions that demand both precise visual details from high-resolution scenes and external knowledge verification."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2602.00593/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"}