Evaluating Parameter Efficient Methods for RLVR

TL;DR

This paper systematically evaluates various parameter-efficient fine-tuning methods for reinforcement learning with verifiable rewards, revealing that structural variants outperform standard approaches and highlighting issues with spectral initialization strategies.

Contribution

It provides the first comprehensive comparison of over 12 PEFT methods in RLVR, identifying superior architectures and analyzing failure modes of common strategies.

Findings

Structural variants like DoRA, AdaLoRA outperform LoRA.

Spectral collapse occurs in SVD-informed initialization methods.

Extreme parameter reduction hampers reasoning capacity.

Abstract

We systematically evaluate Parameter-Efficient Fine-Tuning (PEFT) methods under the paradigm of Reinforcement Learning with Verifiable Rewards (RLVR). RLVR incentivizes language models to enhance their reasoning capabilities through verifiable feedback; however, while methods like LoRA are commonly used, the optimal PEFT architecture for RLVR remains unidentified. In this work, we conduct the first comprehensive evaluation of over 12 PEFT methodologies across the DeepSeek-R1-Distill families on mathematical reasoning benchmarks. Our empirical results challenge the default adoption of standard LoRA with three main findings. First, we demonstrate that structural variants, such as DoRA, AdaLoRA, and MiSS, consistently outperform LoRA. Second, we uncover a spectral collapse phenomenon in SVD-informed initialization strategies (\textit{e.g.,} PiSSA, MiLoRA), attributing their failure to a fundamental misalignment between principal-component updates and RL optimization. Furthermore, our ablations reveal that extreme parameter reduction (\textit{e.g.,} VeRA, Rank-1) severely bottlenecks reasoning capacity. We further conduct ablation studies and scaling experiments to validate our findings. This work provides a definitive guide for advocating for more exploration for parameter-efficient RL methods.