Outcome Accuracy is Not Enough: Aligning the Reasoning Process of Reward Models
This addresses a critical issue in reinforcement learning from human feedback for AI alignment, offering a solution to improve generalization and reduce deceptive behavior in reward models.
The paper tackles the problem of deceptive alignment in generative reward models and LLM-as-a-judge systems, where models produce correct judgments for incorrect reasons, by introducing rationale consistency to align reasoning processes with human judgment. The method achieves state-of-the-art performance with 87.1% on RM-Bench and 82% on JudgeBench, and improves creative writing tasks by 7% in RLHF.
Generative Reward Models (GenRMs) and LLM-as-a-Judge exhibit deceptive alignment by producing correct judgments for incorrect reasons, as they are trained and evaluated to prioritize Outcome Accuracy, which undermines their ability to generalize during RLHF. We introduce Rationale Consistency, a fine-grained metric that quantifies the alignment between the model's reasoning process and human judgment. Our evaluation of frontier models reveals that rationale consistency effectively discriminates among state-of-the-art models and detects deceptive alignment, while outcome accuracy falls short in both respects. To mitigate this gap, we introduce a hybrid signal that combines rationale consistency with outcome accuracy for GenRM training. Our training method achieves state-of-the-art performance on RM-Bench (87.1%) and JudgeBench (82%), surpassing outcome-only baselines by an average of 5%. Using RM during RLHF, our method effectively improves performance as demonstrated on Arena Hard v2, notably yielding a 7% improvement in creative writing tasks. Further analysis confirms that our method escapes the deceptive alignment trap, effectively reversing the decline in rationale consistency observed in outcome-only training.