KernelBench

KernelBench A benchmark designed to evaluate the ability of LLMs to generate efficient GPU kernels for optimizing neural network performance Version [07-21-2025] This HF dataset version has been updated to v0.1 Citation @misc{ouyang2024kernelbench, title={KernelBench: Can LLMs Write GPU Kernels?}, author={Anne Ouyang a

Papers using KernelBench (10)

MusaCoder: Native GPU Kernel Generation with Full-Stack Training on Moore Threads GPU2026TritonRL: Training LLMs to Think and Code Triton Without Cheating2025AutoTriton: Automatic Triton Programming with Reinforcement Learning in LLMs2025CUDA-L1: Improving CUDA Optimization via Contrastive Reinforcement Learning2025 · 1 citesMaxCode: A Max-Reward Reinforcement Learning Framework for Automated Code Optimization2026Dr. Kernel: Reinforcement Learning Done Right for Triton Kernel Generations2026Surprisal-Guided Selection: Compute-Optimal Test-Time Strategies for Execution-Grounded Code Generation2026Fine-Tuning GPT-5 for GPU Kernel Generation2026CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation2026StitchCUDA: An Automated Multi-Agents End-to-End GPU Programing Framework with Rubric-based Agentic Reinforcement Learning2026