CoBA-RL: Capability-Oriented Budget Allocation for Reinforcement Learning in LLMs

TL;DR

CoBA-RL introduces a capability-oriented adaptive budget allocation method for reinforcement learning in LLMs, improving resource efficiency and generalization by dynamically focusing on high-value training samples.

Contribution

It proposes a novel adaptive reinforcement learning algorithm that allocates resources based on model capability, enhancing training efficiency and performance in LLMs.

Findings

Improves resource efficiency in RL for LLMs.

Enhances generalization across multiple benchmarks.

Effectively balances exploration and exploitation.

Abstract

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a key approach for enhancing LLM reasoning. However, standard frameworks like Group Relative Policy Optimization (GRPO) typically employ a uniform rollout budget, leading to resource inefficiency. Moreover, existing adaptive methods often rely on instance-level metrics, such as task pass rates, failing to capture the model's dynamic learning state. To address these limitations, we propose CoBA-RL, a reinforcement learning algorithm designed to adaptively allocate rollout budgets based on the model's evolving capability. Specifically, CoBA-RL utilizes a Capability-Oriented Value function to map tasks to their potential training gains and employs a heap-based greedy strategy to efficiently self-calibrate the distribution of computational resources to samples with high training value. Extensive experiments demonstrate that our approach effectively orchestrates the trade-off between exploration and exploitation, delivering consistent generalization improvements across multiple challenging benchmarks. These findings underscore that quantifying sample training value and optimizing budget allocation are pivotal for advancing LLM post-training efficiency.