Infatoshi/kernelbench-hard-runs
收藏资源简介:
KernelBench-Hard代理运行数据集包含84个完整代理转录本(12个前沿模型×7个问题),这些数据来自在单个Blackwell GPU(RTX PRO 6000, sm_120, CUDA 13.2)上进行的KernelBench-Hard扫描。每个运行包含模型的完整推理轨迹、每个工具调用、最终的solution.py和评估结果。数据集还包括leaderboard.json、problem_baselines.json、annotations/等文件,提供了详细的模型性能、问题基准和注释信息。
The KernelBench-Hard Agent Runs dataset contains 84 full agent transcripts (12 frontier models × 7 problems) from the KernelBench-Hard sweep on a single Blackwell GPU (RTX PRO 6000, sm_120, CUDA 13.2). Each run contains the models full reasoning trace, every tool call, the final `solution.py`, and the eval result. The dataset also includes leaderboard.json, problem_baselines.json, annotations/, and other files, providing detailed model performance, problem benchmarks, and annotation information.




