monitors4codegen
What it is
This repository is the official code and data artifact for the NeurIPS 2023 paper on Monitor-Guided Decoding (MGD). It provides datasets (PragmaticCode, DotPrompts), evaluation scripts, and an implementation of static-analysis monitors that guide code language models to generate compilable code.
The problem it solves
Code language models often hallucinate identifiers or types that do not exist in the repository context, leading to code that fails to compile.
What you could use it for
- Evaluating code language models on repository-level context usage
- Improving compilation rates of generated code via static analysis monitoring
- Reproducing results from the Monitor-Guided Decoding paper
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.