KernelBench: Can LLMs Write Efficient GPU Kernels?

作者：Anne Ouyang, Simon Guo, Simran Arora, Alex Zhang, Wenhua Hu, Christopher Ré, Azalia Mirhoseini · 发表于：arXiv (Cornell University) · 年份：2025 · DOI：10.48550/arxiv.2502.10517 · 被引用次数：1 · 研究领域：Machine Learning and Data Classification、Advanced Neural Network Applications、Image Processing and 3D Reconstruction

Efficient GPU kernels are crucial for building performant machine learning architectures, but writing them is a time-consuming challenge that requires significant expertise; therefore, we explore using language models (LMs) to automate kernel generation. We introduce KernelBench, an open-source framework for evaluating LMs' ability to write fast and correct kernels on a suite of 250 carefully selected PyTorch ML workloads. KernelBench represents a real-world engineering environment and making progress on the introduced benchmark directly translates to faster practical kernels. We introduce a new evaluation metric fast_p, which measures the percentage of generated kernels that are functionally correct and offer a speedup greater than an adjustable threshold p over baseline. Our experiments across various state-of-the-art models and test-time methods show that frontier reasoning models perform the best out of the box but still fall short overall, matching the PyTorch baseline in less than 20% of the cases. While we show that results can improve by leveraging execution and profiling feedback during iterative refinement, KernelBench remains a challenging benchmark, with its difficulty increasing as we raise speedup threshold p.