R$^2$PO: Decoupling Training Trajectories from Inference Responses for LLM Reasoning

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Wang, Jingchu, Xu, Bingbing, Yuan, Yige, Xie, Bin, Sun, Xiaoqian, Shen, Huawei
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866909998660976640
author Wang, Jingchu
Xu, Bingbing
Yuan, Yige
Xie, Bin
Sun, Xiaoqian
Shen, Huawei
author_facet Wang, Jingchu
Xu, Bingbing
Yuan, Yige
Xie, Bin
Sun, Xiaoqian
Shen, Huawei
contents Reinforcement learning has become a central paradigm for improving LLM reasoning. However, existing methods use a single policy to produce both inference responses and training optimization trajectories. The objective conflict between generating stable inference responses and diverse training trajectories leads to insufficient exploration, which harms reasoning capability. In this paper, to address the problem, we propose R$^2$PO (Residual Rollout Policy Optimization), which introduces a lightweight Residual Rollout-Head atop the policy to decouple training trajectories from inference responses, enabling controlled trajectory diversification during training while keeping inference generation stable. Experiments across multiple benchmarks show that our method consistently outperforms baselines, achieving average accuracy gains of 3.4% on MATH-500 and 1.3% on APPS, while also reducing formatting errors and mitigating length bias for stable optimization. Our code is publicly available at https://github.com/RRPO-ARR/Code.
format Preprint
id arxiv_https___arxiv_org_abs_2601_11960
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle R$^2$PO: Decoupling Training Trajectories from Inference Responses for LLM Reasoning
Wang, Jingchu
Xu, Bingbing
Yuan, Yige
Xie, Bin
Sun, Xiaoqian
Shen, Huawei
Machine Learning
Artificial Intelligence
Computation and Language
Reinforcement learning has become a central paradigm for improving LLM reasoning. However, existing methods use a single policy to produce both inference responses and training optimization trajectories. The objective conflict between generating stable inference responses and diverse training trajectories leads to insufficient exploration, which harms reasoning capability. In this paper, to address the problem, we propose R$^2$PO (Residual Rollout Policy Optimization), which introduces a lightweight Residual Rollout-Head atop the policy to decouple training trajectories from inference responses, enabling controlled trajectory diversification during training while keeping inference generation stable. Experiments across multiple benchmarks show that our method consistently outperforms baselines, achieving average accuracy gains of 3.4% on MATH-500 and 1.3% on APPS, while also reducing formatting errors and mitigating length bias for stable optimization. Our code is publicly available at https://github.com/RRPO-ARR/Code.
title R$^2$PO: Decoupling Training Trajectories from Inference Responses for LLM Reasoning
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2601.11960