Making Bias Non-Predictive: Training Robust LLM Reasoning via Reinforcement Learning

ArXi:2602.01528v2 Announce Type: replace-cross Large language models (LLMs) increasingly serve as reasoners and automated evaluators, yet they remain susceptible to cognitive biases -- often altering their reasoning when faced with spurious prompt-level cues such as consensus claims or authority appeals.} Existing mitigations via prompting or supervised fine-tuning fail to generalize, as they modify surface behavior without changing the optimization objective that makes bias cues attractive. We propose \textbf{Epistemic Independence.