TheoPhysEval：GPT-6 Astra 解出 50 道未发表物理难题中的 49 道
新基准显示 AI agent 能解 98% 研究级理论物理题，其中 GPT-6 Astra 在 50 道未发表题上答对 49 道，单题成本低于 3 美元。
Physics is going to be cooked by AI, same as mathematics.🤯 We have never seen an era like this. A new study shows that AI agents can solve 98% of research-level theoretical physics problems, as per new TheoPhysEval benchmark. Even, researchers just tested AI agents on 50 unpublished theoretical physics problems. And GPT-6 Astra solved 49 of them, achieving 98% accuracy at an average reported cost of less than $3 per problem. And if these capabilities continue improving, the pace of scientific discovery could look very different by 2030. we're only beginning to explore what's possible bro! paper: https://arxiv.org/abs/2610.16733
这条是引用着另一条写的：x.com/pratikrath15/status/2110225213401206892