Pinned Loading

1. untell untell Public

   AI-detector auditing toolkit: measures how often a detector flags genuine human writing, how stable its verdict is across seeds, and whether that verdict survives meaning-preserving edits. Detector…

   Python 17 3

2. Scoring-Bias-in-LLM-as-a-Judge Scoring-Bias-in-LLM-as-a-Judge Public

   Scoring bias in LLM-as-a-Judge models: 31 model variants, 3 perturbation probes, 40,500+ judgments. First systematic base vs instruct comparison.

   Python