Prefix Grouper: Efficient GRPO Training through Shared-Prefix Forward

TL;DR

Prefix Grouper significantly reduces the computational overhead of Group Relative Policy Optimization by sharing prefix encoding, enabling scalable and efficient training for long-context scenarios without altering the original optimization dynamics.

Contribution

It introduces a Shared-Prefix Forward strategy that restructures self-attention to eliminate redundant prefix encoding in GRPO, maintaining training equivalence and improving scalability.

Findings

Achieves identical forward outputs and gradients as standard GRPO.

Reduces training computational cost in long-prefix scenarios.

Enables larger group sizes within the same computational budget.

Abstract

Group Relative Policy Optimization (GRPO) enhances policy learning by computing gradients from relative comparisons among candidate outputs that share a common input prefix. Despite its effectiveness, GRPO introduces substantial computational overhead when processing long shared prefixes, which must be redundantly encoded for each group member. This inefficiency becomes a major scalability bottleneck in long-context learning scenarios. We propose Prefix Grouper, an efficient GRPO training algorithm that eliminates redundant prefix computation via a Shared-Prefix Forward strategy. In particular, by restructuring self-attention into two parts, our method enables the shared prefix to be encoded only once, while preserving full differentiability and compatibility with end-to-end training. We provide both theoretical and empirical evidence that Prefix Grouper is training-equivalent to standard GRPO: it yields identical forward outputs and backward gradients, ensuring that the optimization dynamics and final policy performance remain unchanged. Empirically, our experiments confirm that Prefix Grouper achieves consistent results while significantly reducing the computational cost of training, particularly in long-prefix scenarios. The proposed method is fully plug-and-play: it is compatible with existing GRPO-based architectures and can be seamlessly integrated into current training pipelines as a drop-in replacement, requiring no structural modifications and only minimal changes to input construction and attention computation. Prefix Grouper enables the use of larger group sizes under the same computational budget, thereby improving the scalability of GRPO to more complex tasks and larger models. Code is now available at https://github.com/johncaged/PrefixGrouper