Paper

Correct First, Fast Second: What a Local 7B Code Model Can and Cannot Do for GPU Kernel Optimization · ByteTrail