2026
Eliminating Inductive Bias in Reward Models with Information-Theoretic Guidance
ICLR 2026poster
Reward models (RMs) are crucial in reinforcement learning from human feedback (RLHF) to align large language models (LLMs) with human values. However, RM training data is commonly recognized as low-quality, always containing preference conflicts and inductive biases, such as response length or speak…