This page is dedicated to our research on LLM query correction with NSAI.
To-date, we have identified conditions, both domain specific and domain agnostic for LLM’s. Our current work deals with the use of “diversity measures” as a domain independent proxy for detecting errors in LLM output, particularly for symbolic problems.
To try our software package:
- Type pip install langdiversity
- Visit https://github.com/lab-v2/langdiversity
- Read the paper: https://arxiv.org/abs/2308.11189
You may also want to check out our ChatGPT MWP Challenge