Expert-Space Exploration in MoE Reinforcement Learning

• URL: http://arxiv.org/abs/2609.13058v1
• Date: Fri, 11 Sep 2026 16:58:41 GMT
• Title: Expert-Space Exploration in MoE Reinforcement Learning
• Abstract summary: Expert-Space Exploration Reinforcement Learning (ESRL) is an architecture-aware framework that explicitly explores the expert-routing space of MoE models.<n>ESRL achieves the best performance across MoE backbones with top-K, top-1, and shared-expert routing, as well as across mathematics, science, and code tasks without additional sampling or computational cost.
• Score: 41.1632714379556
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Reinforcement learning (RL) has become central to post-training of large language models. Recent advances in RL for Mixture-of-Experts (MoE) models have primarily focused on improving optimization stability and training efficiency, while treating the expert selection as a fixed component. Since routing determines the sparse computation paths that induce output distributions, expert selection offers an additional source of rollout diversity. Through empirical analysis, we find that perturbing expert routing effectively alters model output and increases rollout diversity, which is similar to increasing the decoding temperature. However, direct perturbation can activate unsuitable experts and substantially degrade rollout quality. Motivated by these observations, we introduce Expert-Space Exploration Reinforcement Learning (ESRL), an architecture-aware framework that explicitly explores the expert-routing space of MoE models. ESRL preserves high-confidence experts as anchors, and restricts stochastic routing to a plausible candidate pool, thereby retaining reliable computation paths. The perturbation strength is further adapted according to router entropy to avoid over-perturbation. To mitigate the routing mismatch introduced by perturbation, ESRL records the expert paths used during rollout and replays them during policy optimization. Experiments demonstrate that ESRL achieves the best performance across MoE backbones with top-K, top-1, and shared-expert routing, as well as across mathematics, science, and code tasks without additional sampling or computational cost. Specifically, ESRL on Qwen3-30B-A3B achieves the best among all compared methods, improving average Pass@1 and Pass@8 over GRPO by 3.2 and 4.5 percentage points, respectively. Further analyses of expert utilization and training dynamics provide insights into how exploiting MoE-specific routing structure benefits RL training.

Related papers

• SPS: Steering Probability Squeezing for Better Exploration in Reinforcement Learning for Large Language Models [51.99912169291891]
  Reinforcement learning (RL) has emerged as a promising paradigm for training reasoning-oriented models.<n>We propose a training paradigm that interleaves conventional RL with inverse reinforcement learning (IRL)<n>Our findings suggest that alternating between RL and IRL offers an effective pathway toward extending the exploration capacity of reasoning-oriented large language models.
  arXiv  Detail & Related papers  (2026-04-18T13:49:47Z)
• E3-TIR: Enhanced Experience Exploitation for Tool-Integrated Reasoning [26.716382486149453]
  We propose E3-TIR (Enhanced Experience Exploitation), a warm-up paradigm for the early stages of agent training.<n> Specifically, we formulate training as the dynamic integration of three experience types: Expert Prefixes, Expert Guided, and Self-Exploration.<n>Our method dynamically adapts the model's knowledge boundaries, effectively balancing exploration diversity with training efficiency.
  arXiv  Detail & Related papers  (2026-04-10T16:14:48Z)
• RetroAgent: From Solving to Evolving via Retrospective Dual Intrinsic Feedback [54.39884046754265]
  RetroAgent is an online RL framework that enables agents to master complex interactive environments.<n>Our experiments show that RetroAgent achieves state-of-the-art (SOTA) performance.
  arXiv  Detail & Related papers  (2026-03-09T16:23:33Z)
• Towards Stable and Effective Reinforcement Learning for Mixture-of-Experts [113.0656076371565]
  We propose a novel router-aware approach to optimize importance sampling weights in off-policy reinforcement learning (RL)<n>Specifically, we design a rescaling strategy guided by router logits, which effectively reduces gradient variance and mitigates training divergence.<n> Experimental results demonstrate that our method significantly improves both the convergence stability and the final performance of MoE models.
  arXiv  Detail & Related papers  (2025-10-27T05:47:48Z)
• Stabilizing MoE Reinforcement Learning by Aligning Training and Inference Routers [33.18639109838351]
  We analyze the training-inference consistency of MoE models and identify a notable discrepancy in routing behaviors.<n>We propose Rollout Replay Routing (R3), a method that records routing distributions from the inference engine and replays them during training.<n>R3 significantly reduces training-inference policy KL divergence and mitigates extreme discrepancies without compromising training speed.
  arXiv  Detail & Related papers  (2025-10-13T13:11:27Z)
• Reinforcement Learning on Pre-Training Data [55.570379963147424]
  We introduce Reinforcement Learning on Pre-Training data (R), a new training-time scaling paradigm for optimizing large language models (LLMs)<n>R enables the policy to autonomously explore meaningful trajectories to learn from pre-training data and improve its capability through reinforcement learning (RL)<n>Extensive experiments on both general-domain and mathematical reasoning benchmarks across multiple models validate the effectiveness of R.
  arXiv  Detail & Related papers  (2025-09-23T17:10:40Z)
• Agentic Reinforcement Learning with Implicit Step Rewards [92.26560379363492]
  Large language models (LLMs) are increasingly developed as autonomous agents using reinforcement learning (agentic RL)<n>We introduce implicit step rewards for agentic RL (iStar), a general credit-assignment strategy that integrates seamlessly with standard RL algorithms.<n>We evaluate our method on three challenging agent benchmarks, including WebShop and VisualSokoban, as well as open-ended social interactions with unverifiable rewards in SOTOPIA.
  arXiv  Detail & Related papers  (2025-09-23T16:15:42Z)
• Beyond Accuracy: Dissecting Mathematical Reasoning for LLMs Under Reinforcement Learning [93.00629872970364]
  Reinforcement learning (RL) has become the dominant paradigm for improving the performance of language models on complex reasoning tasks.<n>We introduce SPARKLE, a fine-grained analytic framework to dissect the effects of RL across three key dimensions.<n>We study whether difficult problems -- those yielding no RL signals and mixed-quality reasoning traces -- can still be effectively used for training.
  arXiv  Detail & Related papers  (2025-06-05T07:53:59Z)
• DRL4Route: A Deep Reinforcement Learning Framework for Pick-up and Delivery Route Prediction [21.335721424944257]
  We present the first attempt to generalize Reinforcement Learning (RL) to the route prediction task, leading to a novel RL-based framework called DRL4Route. DRL4Route can serve as a plug-and-play component to boost the existing deep learning models. It follows the actor-critic architecture which is equipped with a Generalized Advantage Estimator.
  arXiv  Detail & Related papers  (2023-07-30T14:50:31Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.