Mistrial I Re run your LLM judge and see how often it flips

AI · Practice · 2026-10-27 00:46

Mistrial is a GitHub tool that repeatedly runs an LLM judge over the same set of samples and measures how often its verdicts flip, helping assess the stability and reliability of AI evaluation results.

Source: github.com

Auto-generated summary for reference only. RSS

Mistrial I Re run your LLM judge and see how often it flips

Mistrial I Re run your LLM judge and see how often it flips