Reinforcement Learning Finetunes Small Subnetworks in Large Language Models

TL;DR

Reinforcement learning fine-tunes only a small, sparse subnetwork within large language models, achieving comparable performance to full fine-tuning without explicit sparsity constraints.

Contribution

This study reveals that RL fine-tuning induces intrinsic parameter update sparsity across diverse models and algorithms, with small subnetworks capturing most performance gains.

Findings

RL fine-tuning updates only 5-30% of parameters.

Subnetwork fine-tuning recovers full test accuracy.

Update sparsity is consistent across models and algorithms.

Abstract

Reinforcement learning (RL) yields substantial improvements in large language models (LLMs) downstream task performance and alignment with human values. Surprisingly, such large gains result from updating only a small subnetwork comprising just 5 percent to 30 percent of the parameters, with the rest effectively unchanged. We refer to this phenomenon as parameter update sparsity induced by RL. It is observed across all 7 widely used RL algorithms (e.g., PPO, GRPO, DPO) and all 10 LLMs from different families in our experiments. This sparsity is intrinsic and occurs without any explicit sparsity promoting regularizations or architectural constraints. Finetuning the subnetwork alone recovers the test accuracy, and, remarkably, produces a model nearly identical to the one obtained via full finetuning. The subnetworks from different random seeds, training data, and even RL algorithms show substantially greater overlap than expected by chance. Our analysis suggests that this sparsity is not due to updating only a subset of layers, instead, nearly all parameter matrices receive similarly sparse updates. Moreover, the updates to almost all parameter matrices are nearly full-rank, suggesting RL updates a small subset of parameters that nevertheless span almost the full subspaces that the parameter matrices can represent. We conjecture that the this update sparsity can be primarily attributed to training on data that is near the policy distribution, techniques that encourage the policy to remain close to the pretrained model, such as the KL regularization and gradient clipping, have limited impact.