Morning Edition · Monday, August 3, 2026Published at 1:38 AM EDT · New York

Paper Proposes Cross-Model Auditing to Harden LLM Judges Against Their Own Biases

Chain-of-Models routes a judgment through multiple models to identify the cognitive biases that prompt-based debiasing fails to fix.

Language models are increasingly used as automated judges, in evaluation harnesses, in reinforcement-learning reward signals, and in production content grading, but their verdicts inherit cognitive biases such as position and verbosity effe…