What AI says about EvalLM

EvalLM is an open-source, interactive tool for evaluating and comparing prompts for large language models using user-defined natural-language criteria. It offers a three-panel interface (Generation, Data, and Evaluation) to describe shared instructions, build two prompt templates, upload a JSON dataset of inputs, and compare generated outputs against custom criteria. The project is MIT-licensed on GitHub, with a live demo that requires an OpenAI API key and is associated with a CHI 2024 paper.