Complementing reinforcement learning with SFT through logit averaging in the post training of LLMs
For LLM post-training, this method offers a simpler alternative to KL-regularized RLVR that maintains formatting quality while improving reasoning.
The authors propose a logit averaging method that combines a frozen SFT policy with a trainable policy in GRPO, eliminating the need for KL regularization or a critic. On MATH, cn-k12, and MMLU, it achieves higher or comparable accuracy to KL-regularized GRPO.
We introduce a novel method that averages the logits of a frozen reference policy (e.g., SFT) and a trainable policy, and incorporate the method into Group Relative Policy Optimization (GRPO). In contrast to Reinforcement Learning with Verifiable Rewards (RLVR) methods, our proposal does not involve a Kullback Leibler (KL) regularization or critic; the trainable policy and the reference anchor are coupled through the logit averaging structure to leverage the reasoning expertise of the trainable policy while maintaining the formatting advantage of SFT. Our method is evaluated on MATH, cn-k12, and MMLU, and the results show a higher accuracy or at least comparable accuracy relative to the canonical KL-regularized GRPO.