Reward Models Identify Consistency, Not Causality

TL;DR

This paper reveals that reward models for language models focus more on structural consistency and reasoning patterns than on actual causal correctness, highlighting a key limitation in current alignment techniques.

Contribution

The study demonstrates that reward models prioritize consistency over causality, challenging assumptions and suggesting the need for causality-aware reward modeling approaches.

Findings

Reward models rely heavily on structural consistency.

Removing problem statements minimally affects reward scores.

Disrupting reasoning flow significantly impacts reward outputs.

Abstract

Reward models (RMs) play a crucial role in aligning large language models (LLMs) with human preferences and enhancing reasoning quality. Traditionally, RMs are trained to rank candidate outputs based on their correctness and coherence. However, in this work, we present several surprising findings that challenge common assumptions about RM behavior. Our analysis reveals that state-of-the-art reward models prioritize structural consistency over causal correctness. Specifically, removing the problem statement has minimal impact on reward scores, whereas altering numerical values or disrupting the reasoning flow significantly affects RM outputs. Furthermore, RMs exhibit a strong dependence on complete reasoning trajectories truncated or incomplete steps lead to significant variations in reward assignments, indicating that RMs primarily rely on learned reasoning patterns rather than explicit problem comprehension. These findings hold across multiple architectures, datasets, and tasks, leading to three key insights: (1) RMs primarily assess coherence rather than true reasoning quality; (2) The role of explicit problem comprehension in reward assignment is overstated; (3) Current RMs may be more effective at ranking responses than verifying logical validity. Our results suggest a fundamental limitation in existing reward modeling approaches, emphasizing the need for a shift toward causality-aware reward models that go beyond consistency-driven evaluation.