Shubham Jha

Functorial Lowering for Verifiable Cross-Platform Code Generation with Large Language Models: A Position Paper

Large language models (LLMs) generate plausible code but offer no guarantee that it is correct or that it exploits the target hardware well. Compilers offer the opposite trade-off: rigorous, structure-preserving translation, but only along optimizations their authors anticipated. We argue that category theory provides...