GPT-6.1 Sol examined nearly 800 mathematics papers, finding major errors in 6% and minor mistakes in another 30%. Researchers estimate 90% of its major findings are correct. As AI adoption surged, the major-error rate in newer papers fell to just 1.5%.
Summary
GPT-6.1 Sol 审查了近800篇数学论文，发现6%存在重大错误、30%有小问题，且随着AI采用率上升，新论文的重大错误率降至1.5%。
Similar Articles
GPT-6 Astra autonomously solves 49 out of 50 unpublished theoretical physics research problems, scoring 98% at under $3 per problem. Researchers initially marked several answers wrong, only to discover the AI had caught errors in the calculations of world-class physicists.
OpenAI的GPT-6 Astra在未公开的理论物理研究问题上自主解决了50题中的49题，准确率达98%，每题成本低于3美元，甚至发现了世界顶级物理学家计算中的错误。
Claude Opus 5.5 and GPT-6 Astra independently audit 138 economics papers, flagging 2,035 potential problems and 147 potentially headline-changing findings. A third AI attempts to verify and refute them, fueling debate over AI systematically re-examining decades of published research.
Claude Opus 5.5 and GPT-6 Astra independently audited 138 economics papers, flagging 2,035 potential problems and 147 potentially headline-changing findings, while a third AI attempted verification, sparking debate over AI systematically re-examining decades of published research.
ChatGPT Sol 5.6 high found a normalization error in two recently published Riemann Hypothesis papers. The author confirmed it.
A non-mathematician used ChatGPT to identify a normalization error in two recently published Riemann Hypothesis papers, and the author confirmed the issue after being contacted. The story highlights AI's growing role in assisting mathematical research.
On SWEBench Pro, 68.5% of GPT 5.5’s failures were caused by broken or incorrect test cases, totaling 28.9% of the entire benchmark
An analysis reveals that 28.9% of GPT 5.5's failures on SWEBench Pro are due to broken or incorrect test cases, and similar issues affect other major AI benchmarks, raising concerns about the accuracy of current evaluation methods.
GPT 5.6 Sol benchmarks
GPT 5.6 Sol achieves new benchmark results, showcasing performance improvements in AI language modeling.