谢赛宁·· 2025-06-17精选
So this is not a benchmark for software engineering agents. It’s meant to test core reasoning and in...
So this is not a benchmark for software engineering agents. It’s meant to test core reasoning and in...
AI 导读
所以这不是一个针对软件工程智能体的基准测试。它旨在通过编程测试核心推理与智能——由一些顶尖竞技程序员撰写的 71 页深度分析作为支撑。
推荐理由
o3与Gemini 2.5在IOI级竞赛题上零分，LLM推理天花板显现
来源：谢赛宁 · x.com