Recently, there have been a few interesting papers that essentially treat LLMs as a compilers - most notably KernelBench and Sakana’s AI CUDA Engineer (though note that the Sakana paper had some issues around the model reward hacking).
Share this post
LLMs as compilers
Share this post
Recently, there have been a few interesting papers that essentially treat LLMs as a compilers - most notably KernelBench and Sakana’s AI CUDA Engineer (though note that the Sakana paper had some issues around the model reward hacking).