Monitorability as a Free Gift: How RLVR Spontaneously Aligns Reasoning

TL;DR

This paper investigates how monitorability, the reflection of internal reasoning in chain-of-thought traces, can spontaneously improve during RLVR training, influenced by data diversity and training dynamics, but not necessarily linked to reasoning performance.

Contribution

It provides a systematic evaluation of monitorability emergence in RLVR, highlighting data dependence, its independence from reasoning capability, and mechanistic insights into its underlying factors.

Findings

Monitorability improvements are data-dependent.

Data diversity and instruction-following data are critical.

Monitorability is orthogonal to reasoning performance.

Abstract

As Large Reasoning Models (LRMs) are increasingly deployed, auditing their chain-of-thought (CoT) traces for safety becomes critical. Recent work has reported that monitorability--the degree to which CoT faithfully and informatively reflects internal computation--can appear as a "free gift" during the early stages of Reinforcement Learning with Verifiable Rewards (RLVR). We make this observation concrete through a systematic evaluation across model families and training domains. Our results show that this effect is not universal: monitorability improvements are strongly data-dependent. In particular, we demonstrate the critical role of data diversity and instruction-following data during RLVR training. We further show that monitorability is orthogonal to capability--improvements in reasoning performance do not imply increased transparency. Through mechanistic analysis, we attribute monitorability gains primarily to response distribution sharpening (entropy reduction) and increased attention to the prompt, rather than stronger causal reliance on reasoning traces. We also reveal how monitorability dynamics vary with controlled training and evaluation difficulty. Together, these findings provide a holistic view of how monitorability emerges under RLVR, clarifying when gains are likely to occur and when they are not.