{"paper":{"title":"Measuring short-form factuality in large language models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"SimpleQA benchmark measures if language models know what they know on short facts.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Amelia Glaese, Hyung Won Chung, Jason Wei, John Schulman, Nguyen Karina, Spencer Papay, William Fedus, Yunxin Joy Jiao","submitted_at":"2024-11-07T01:58:42Z","abstract_excerpt":"We present SimpleQA, a benchmark that evaluates the ability of language models to answer short, fact-seeking questions. We prioritized two properties in designing this eval. First, SimpleQA is challenging, as it is adversarially collected against GPT-4 responses. Second, responses are easy to grade, because questions are created such that there exists only a single, indisputable answer. Each answer in SimpleQA is graded as either correct, incorrect, or not attempted. A model with ideal behavior would get as many questions correct as possible while not attempting the questions for which it is n"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"SimpleQA is a simple, targeted evaluation for whether models 'know what they know,' and our hope is that this benchmark will remain relevant for the next few generations of frontier models.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"Questions can be created such that there exists only a single, indisputable answer and that adversarial collection against GPT-4 responses produces questions that remain challenging for future models.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SimpleQA is a new benchmark of short, single-answer factual questions collected adversarially against GPT-4 to evaluate LLM factuality and confidence calibration.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"SimpleQA benchmark measures if language models know what they know on short facts.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f90715c97f4bf0896a9b640e57e6cbc403e21e7ba9b86a45a790a8bd7c60c7fa"},"source":{"id":"2411.04368","kind":"arxiv","version":1},"verdict":{"id":"d603a5c1-f3a8-4bf9-81c8-f9976ebdcd58","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T06:41:33.689500Z","strongest_claim":"SimpleQA is a simple, targeted evaluation for whether models 'know what they know,' and our hope is that this benchmark will remain relevant for the next few generations of frontier models.","one_line_summary":"SimpleQA is a new benchmark of short, single-answer factual questions collected adversarially against GPT-4 to evaluate LLM factuality and confidence calibration.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"Questions can be created such that there exists only a single, indisputable answer and that adversarial collection against GPT-4 responses produces questions that remain challenging for future models.","pith_extraction_headline":"SimpleQA benchmark measures if language models know what they know on short facts."},"references":{"count":19,"sample":[{"doi":"","year":2024,"title":"org/abs/2305.18248","work_id":"9eeaa979-72cb-4989-b9c7-8b50900f3647","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"P. Anthropic. Claude 3 model card, 2024. URL https://www-cdn.anthropic.com/de8ba9b01c9ab7cbabf5c33b80b7bbc618857627/Model_Card_Claude_3.pdf","work_id":"613d427d-70ca-468a-a12f-43da8b7df476","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"In Proceedings of the 22nd international conference on Machine learning, pages 89–96","work_id":"27f57922-254f-4edb-9365-44ba354e34b2","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2017,"title":"TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension","work_id":"f20e62ba-6265-4b97-aa8c-ddefaf2f5762","ref_index":4,"cited_arxiv_id":"1705.03551","is_internal_anchor":true},{"doi":"","year":2022,"title":"Language Models (Mostly) Know What They Know","work_id":"8ca58a10-da41-4f70-baae-7e449512e345","ref_index":5,"cited_arxiv_id":"2207.05221","is_internal_anchor":true}],"resolved_work":19,"snapshot_sha256":"a63a5b722d18f18394072fab950867951a3b3b95e5eb6d9f5d19859c8a0de3e0","internal_anchors":4},"formal_canon":{"evidence_count":3,"snapshot_sha256":"7f1c924b744a9cf07262c238dd2aaf290fe366799f332ffd18b4278061cdd1ce"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}