Does LLM Alignment Really Need Diversity? An Empirical Study of Adapting RLVR Methods for Moral Reasoning

TL;DR

This study empirically compares distribution-matching and reward-maximizing RLVR methods for LLM alignment in moral reasoning, finding that standard reward-maximizing approaches are sufficiently effective without requiring diversity-preserving algorithms.

Contribution

The paper provides the first comprehensive empirical comparison of RLVR paradigms on moral reasoning tasks, challenging the assumption that diversity-seeking methods are necessary for alignment.

Findings

Reward-maximizing methods perform as well or better than distribution-matching approaches.

Moral reasoning responses are more concentrated in semantic space than mathematical reasoning.

Diversity-preserving algorithms are not inherently necessary for effective LLM alignment.

Abstract

Reinforcement learning with verifiable rewards (RLVR) has achieved remarkable success in logical reasoning tasks, yet whether large language model (LLM) alignment requires fundamentally different approaches remains unclear. Given the apparent tolerance for multiple valid responses in moral reasoning, a natural hypothesis is that alignment tasks inherently require diversity-seeking distribution-matching algorithms rather than reward-maximizing policy-based methods. We conduct the first comprehensive empirical study comparing both paradigms on MoReBench. To enable stable RLVR training, we build a rubric-grounded reward pipeline by training a Qwen3-1.7B judge model. Contrary to our hypothesis, we find that distribution-matching approaches do not demonstrate significant advantages over reward-maximizing methods as expected on alignment tasks. Through semantic visualization mapping high-reward responses to semantic space, we demonstrate that moral reasoning exhibits more concentrated high-reward distributions than mathematical reasoning, where diverse solution strategies yield similarly high rewards. This counter-intuitive finding explains why mode-seeking optimization proves equally or more effective for alignment tasks. Our results suggest that alignment tasks do not inherently require diversity-preserving algorithms, and standard reward-maximizing RLVR methods can effectively transfer to moral reasoning without explicit diversity mechanisms.