Perturbation Testing Python Packages

yuragi

yuragi — LLM Confidence Fragility Analyzer. Perturbation-driven hallucination detection with workshop-grade real benchmarks (TruthfulQA n=412 ensemble AUC 0.73, TriviaQA n=200 confidence-inversion AUC 0.75).

yuragi-ai

yuragi — LLM Confidence Fragility Analyzer. Perturbation-driven hallucination detection with workshop-grade real benchmarks (TruthfulQA n=412 ensemble AUC 0.73, TriviaQA n=200 confidence-inversion AUC 0.75).

yuragi-guardrails

yuragi — LLM Confidence Fragility Analyzer. Perturbation-driven hallucination detection with workshop-grade real benchmarks (TruthfulQA n=412 ensemble AUC 0.73, TriviaQA n=200 confidence-inversion AUC 0.75).

falsifyai

Falsification-first reliability testing for AI systems: perturb inputs, preserve replayable evidence, diff reliability across model changes.