Skip to content

Beyond On-Policy Exploration: Integrating External Policy Rollouts for Reinforcement Learning in Diffusion Language Models

Aug 2026 · 0 citations · 43 references
Computer Science

TL;DR

This work proposes External Rollout Integration with Length Control and Source-Specific Processing (ERILS), which controls external-rollout length and processes the rewards of on-policy and external rollouts separately, and shows that length-controlled external rollouts are more effective than uncontrolled external rollouts.

Abstract

Recent reinforcement learning methods for diffusion large language models (dLLMs) commonly rely on on-policy rollouts generated by the target dLLM itself. When successful on-policy rollouts are scarce, however, on-policy training may receive little positive reward and make only limited progress. To mitigate this problem, we explore incorporating higher-reward rollouts generated by a stronger external policy alongside on-policy rollouts from the target dLLM. However, directly incorporating these external rollouts introduces two practical challenges: differences in rollout length and instability when jointly processing rewards from on-policy and external rollouts. To address these challenges, we propose External Rollout Integration with Length Control and Source-Specific Processing (ERILS), which controls external-rollout length and processes the rewards of on-policy and external rollouts separately. Experiments on Sudoku, Countdown, and MATH500 under zero-shot evaluation show that ERILS improves multi-sample performance across all three tasks, with the largest gains on Sudoku. On Sudoku, ERILS achieves 98.4% best-of-4 completion accuracy, compared with 40.3% for the strongest baseline. ERILS also maintains approximately 90% deterministic single-completion accuracy on Sudoku across generation lengths of 128, 256, and 512 tokens. Our component analysis further shows that length-controlled external rollouts are more effective than uncontrolled external rollouts, and that source-specific reward processing avoids the training collapse observed with joint reward processing. These results show that rollout construction and reward processing are important design dimensions when integrating external rollouts into dLLM reinforcement learning.

View source

Similar papers

#artificial intelligence Review Sep 2026

MInTRL: Off-policy Intervention can boost On-policy RL

This work introduces Minimal Intervention Reinforcement Learning (MInTRL), which expands the exploration frontier through sparse, local interventions in otherwise on-policy rollouts, and establishes minimal intervention as an effective paradigm for enhancing on-policy RL.

Ming-Yu Chen, Ye-Fan Tao, Gerald Friedland et al. · 1 citation

Learning Beyond What You Sample: Off-Policy-Aware Cross-Model Trajectory Exchange for RLVR

GRAFT (Gated Replacement of Answer-Failed groups with peer Trajectories), an off-policy-aware framework that replaces all-fail groups with informative peer groups and transfers both successful and unsuccessful peer responses with peer-computed advantages, while controlling cross-model mismatch through sequence-level co...

Doohyuk Jang, Y. Park, Gyouk Chu et al. · 0 citations
Review Open access Sep 2026

Reinforcement Learning Post-Training for Reasoning Large Language Models: Methods, Systems, and Evaluation

This survey develops a mechanism-driven framework for separating effects of reinforcement learning, and organizes methods along three axes— learning-signal type, policy/data regime, and optimization granularity—and decomposes their training stacks into reusable motifs spanning advantage construction, drift control, den...

Liu Yang, Han Zhu, Zheng-Yang Zhong et al. · 1 citation

SIPO: Unifying Reinforcement Learning with On-Policy Self-Distillation

This work proposes self-instructing policy optimization (SIPO) with a contrastive self-teacher to provide dense credit, and demonstrates that SIPO outperforms both RLVR and OPSD baselines without an external teacher or additional generation.

Zhenrui Yue, Hui-Min Zeng, Yue-Qi Wang et al. · 1 citation

Related blog posts

MIT News · Artificial Intelligence Nov 17, 2026

Learning the language of machine learning

By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.

Microsoft Research Blog Nov 16, 2026

TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient

AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.