Lakshya A Agrawal (@LakshyAAAgrawal) on X

X (formerly Twitter) ·

1 min read Original article ↗

Post

Post

Lakshya A Agrawal on X: "LLMs often hallucinate incorrect names, especially in private codebases. We introduce Monitor-Guided Decoding (MGD)-guide LMs to generate compilable code with correct symbol names more reliably!🚀 (Work w/ @adityakanade0 Navin Goyal @LahiriShuvendu @SriramRajamani @MSFTResearch)"

@LakshyAAAgrawal

Example where text-davinci-003 and SantaCoder generate wrong identifiers, but SantaCoder with monitor-guided decoding generates correct identifiers.

@_akhaliq

Guiding Language Models of Code with Global Context using Monitors paper page: huggingface.co/papers/2306.10… Language models of code (LMs) work well when the surrounding code in the vicinity of generation provides sufficient context. This is not true when it becomes necessary to use