joeljang/negated-prompts-for-llms

This project is for researchers and practitioners working with large language models (LLMs) who need to understand how well these models interpret negative statements or instructions. It helps you assess an LLM's comprehension of negation by providing tools to input prompts containing negated concepts and then analyze the model's responses to see if it correctly understands what *not* to do or say. Anyone evaluating LLM robustness or behavior for specific applications would find this useful.

No commits in the last 6 months.