Improving Data Efficiency for LLM Reinforcement Fine-tuning Through Difficulty-targeted Online Data Selection and Rollout Replay

TL;DR

This paper introduces methods to improve data efficiency in reinforcement learning fine-tuning of large language models by using adaptive difficulty-based data selection and rollout replay, significantly reducing training time.

Contribution

It proposes a novel adaptive difficulty estimation framework and rollout replay mechanism to enhance data efficiency in LLM RL fine-tuning, reducing resource consumption.

Findings

Reduces RL fine-tuning time by up to 62%.

Maintains performance levels comparable to existing algorithms.

Efficient difficulty estimation with minimal rollouts.

Abstract

Reinforcement learning (RL) has become an effective approach for fine-tuning large language models (LLMs), particularly to enhance their reasoning capabilities. However, RL fine-tuning remains highly resource-intensive, and existing work has largely overlooked the problem of data efficiency. In this paper, we propose two techniques to improve data efficiency in LLM RL fine-tuning: difficulty-targeted online data selection and rollout replay. We introduce the notion of adaptive difficulty to guide online data selection, prioritizing questions of moderate difficulty that are more likely to yield informative learning signals. To estimate adaptive difficulty efficiently, we develop an attention-based framework that requires rollouts for only a small reference set of questions. The adaptive difficulty of the remaining questions is then estimated based on their similarity to this set. To further reduce rollout cost, we introduce a rollout replay mechanism inspired by experience replay in traditional RL. This technique reuses recent rollouts, lowering per-step computation while maintaining stable updates. Experiments across 6 LLM-dataset combinations show that our method reduces RL fine-tuning time by 23% to 62% while reaching the same level of performance as the original GRPO algorithm. Our code is available at https://github.com/ASTRAL-Group/data-efficient-llm-rl.