explore judgy for correcting LLM judge's bias

[linminhtoo]

https://github.com/ai-evals-course/judgy

When using Large Language Models (LLMs) as judges to evaluate other models or systems, the judge's own biases and errors can significantly impact the reliability of the evaluation. judgy provides tools to estimate the true success rate of your system by correcting for LLM judge bias, and bootstrapping to generate a confidence interval.

[linminhtoo]

note that we already do have a mechanism for manually annotating judge responses to mark if the judge's assessment was correct or not. this can then be used to calculate metrics on the judge's performance, allowing us to iterate the judge prompt/harness to improve these metrics, in order to improve the judge's reliability