Comment on Ask HN: How are you using LLMs for traversing decompiler output?Comments−userbinator1yUnfortunately LLMs are not good at precision and details, which is exactly what you need for the sort of analysis you're trying to do.−menaerus1yRight. Have a look at the paper above from Meta on how they fine-tuned the Code Llama with LLVM IR to beat the compiler in producing size-optimized binaries.
Comments
Unfortunately LLMs are not good at precision and details, which is exactly what you need for the sort of analysis you're trying to do.
Right. Have a look at the paper above from Meta on how they fine-tuned the Code Llama with LLVM IR to beat the compiler in producing size-optimized binaries.