Learning at the Right Pace: Adaptive Data Scheduling Improves LLM Reinforcement Learning

2026-06-20

AI Summary

This work addresses the limitation of uniform sampling in reinforcement learning-based post-training of large language models, which overlooks the semantic structure of data and the dynamic evolution of policy capabilities. To overcome this, the paper proposes an Adaptive Data Scheduling (ADS) framework featuring a novel two-level adaptive mechanism: it dynamically adjusts the sampling distribution across semantic clusters while prioritizing boundary samples within each cluster to enable fine-grained control over the training process. Extensive experiments demonstrate that ADS consistently improves accuracy by an average of 5.2% across three prominent large language models and seven reasoning benchmarks. Moreover, ADS is compatible with diverse reinforcement learning objectives and significantly outperforms existing baseline methods.

Abstract

Large Language Models (LLMs) achieve remarkable reasoning capabilities through reinforcement learning (RL) post-training. However, existing RL post-training commonly relies on uniform data sampling, which ignores the semantic structure of the training data and the changing capability of the training policy. To address these limitations, we propose Adaptive Data Scheduling (ADS), a dual-level data scheduling framework for pacing RL post-training that replaces uniform sampling with an adaptive distribution over semantic clusters and policy-boundary sample selection. At the cluster level, ADS organizes samples according to semantic patterns and maintains an adaptive inter-cluster distribution to solidify current training progress. At the sample level, ADS performs intra-cluster scheduling to continuously sample policy-boundary samples, which provides informative relative advantages. Experimental results across three LLMs and seven reasoning benchmarks demonstrate that ADS improves average accuracy by 5.2% over Group Relative Policy Optimization (GRPO). Notably, ADS consistently improves RL methods with different objective designs, highlighting its potential as a general data scheduling strategy for LLM RL post-training. The source code is available at: https://github.com/Richard-zrx/ADS.

Problem

• reinforcement learning
• data scheduling
• large language models
• semantic structure
• policy adaptation

Innovation

• Adaptive Data Scheduling
• Reinforcement Learning
• Semantic Clustering
• Policy-Boundary Sampling
• LLM Post-Training