Contradish tests whether AI models provide consistent answers when given semantically equivalent prompts rather than just measuri

[prismscholars]

contradish is a python library now being used by legal teams to audit AI reasoning processes. it detects contradictions and returns the CAI score and failure reason