SPEC-RL: Accelerating On-Policy Reinforcement Learning via Speculative Rollouts

• URL: http://arxiv.org/abs/2509.23232v1
• Date: Sat, 27 Sep 2025 10:32:34 GMT
• Title: SPEC-RL: Accelerating On-Policy Reinforcement Learning via Speculative Rollouts
• Authors: Bingshuai Liu, Ante Wang, Zijun Min, Liang Yao, Haibo Zhang, Yang Liu, Anxiang Zeng, Jinsong Su,
• Abstract summary: SPEC-RL is a framework that integrates SPECulative decoding with the RL rollout process.<n>It reduces rollout time by 2-3x without compromising policy quality.<n>As a purely rollout-stage enhancement, SPEC-RL integrates seamlessly with mainstream algorithms.
• Score: 35.82325476805143
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large Language Models (LLMs) increasingly rely on reinforcement learning with verifiable rewards (RLVR) to elicit reliable chain-of-thought reasoning. However, the training process remains bottlenecked by the computationally expensive rollout stage. Existing acceleration methods-such as parallelization, objective- and data-driven modifications, and replay buffers-either incur diminishing returns, introduce bias, or overlook redundancy across iterations. We identify that rollouts from consecutive training epochs frequently share a large portion of overlapping segments, wasting computation. To address this, we propose SPEC-RL, a novel framework that integrates SPECulative decoding with the RL rollout process. SPEC-RL reuses prior trajectory segments as speculative prefixes and extends them via a draft-and-verify mechanism, avoiding redundant generation while ensuring policy consistency. Experiments on diverse math reasoning and generalization benchmarks, including GSM8K, MATH-500, OlympiadBench, MMLU-STEM, and others, demonstrate that SPEC-RL reduces rollout time by 2-3x without compromising policy quality. As a purely rollout-stage enhancement, SPEC-RL integrates seamlessly with mainstream algorithms (e.g., PPO, GRPO, DAPO), offering a general and practical path to scale RLVR for large reasoning models. Our code is available at https://github.com/ShopeeLLM/Spec-RL

Related papers

• Controllable Exploration in Hybrid-Policy RLVR for Multi-Modal Reasoning [88.42566960813438]
  CalibRL is a hybrid-policy RLVR framework that supports controllable exploration with expert guidance.<n>CalibRL increases policy entropy in a guided manner and clarifies the target distribution.<n>Experiments across eight benchmarks, including both in-domain and out-of-domain settings, demonstrate consistent improvements.
  arXiv  Detail & Related papers  (2026-02-22T07:23:36Z)
• Discover, Learn, and Reinforce: Scaling Vision-Language-Action Pretraining with Diverse RL-Generated Trajectories [33.872433985210876]
  Scaling vision-language-action (VLA) model pre-training requires large volumes of diverse, high-quality manipulation trajectories.<n>We propose Discover, Lea rn and Reinforce, which generates multiple distinct, high-success behavioral patterns for VLA pretraining.<n>When adapted to unseen downstream task suites, VLA models pretrained on our diverse RL data surpass counterparts trained on equal-sized standard RL datasets.
  arXiv  Detail & Related papers  (2025-11-24T07:54:49Z)
• Beat the long tail: Distribution-Aware Speculative Decoding for RL Training [75.75462952580796]
  We propose a Distribution Aware Speculative decoding framework that accelerates RL rollouts without altering model outputs.<n>Experiments on math and code reasoning tasks show that DAS reduces rollout time up to 50% while preserving identical training curves.
  arXiv  Detail & Related papers  (2025-11-17T19:02:12Z)
• RLoop: An Self-Improving Framework for Reinforcement Learning with Iterative Policy Initialization [65.23034604711489]
  We introduce RLoop, a self-improving framework for training large reasoning models.<n>RLoop transforms the standard training process into a virtuous cycle: it first uses RL to explore the solution space from a given policy, then filters the successful trajectories to create an expert dataset.<n>Our experiments show RLoops forgetting and substantially improves generalization, boosting average accuracy by 9% and pass@32 by over 15% compared to vanilla RL.
  arXiv  Detail & Related papers  (2025-11-06T11:27:16Z)
• From Supervision to Exploration: What Does Protein Language Model Learn During Reinforcement Learning? [76.288870982181]
  Protein language models (PLMs) have advanced computational protein science through large-scale pretraining and scalable architectures.<n> reinforcement learning (RL) has broadened exploration and enabled precise multi-objective optimization in protein design.<n>We ask if RL improves sampling efficiency and, more importantly, if it reveals capabilities not captured by supervised learning.
  arXiv  Detail & Related papers  (2025-10-02T01:31:10Z)
• APRIL: Active Partial Rollouts in Reinforcement Learning to Tame Long-tail Generation [40.120847511378365]
  Reinforcement learning (RL) has become a cornerstone in advancing large-scale pre-trained language models (LLMs)<n>We propose Active Partial Rollouts in Reinforcement Learning (APRIL), which mitigates long-tail inefficiency.<n>APRIL over-provisions rollout requests, terminates once the target number of responses is reached, and recycles incomplete responses for continuation in future steps.
  arXiv  Detail & Related papers  (2025-09-23T01:32:36Z)
• History Rhymes: Accelerating LLM Reinforcement Learning with RhymeRL [14.506189610798929]
  Reinforcement learning (RL) has emerged as a pivotal methodology for enhancing the reasoning capabilities of large language models (LLMs)<n>We introduce RhymeRL, an LLM RL system designed to accelerate RL training with two key innovations.<n>First, to enhance rollout generation, we present HistoSpec, a speculative decoding inference engine.<n>Second, to tackle rollout bubbles, we introduce HistoPipe, a two-tier scheduling strategy.
  arXiv  Detail & Related papers  (2025-08-26T01:42:46Z)
• 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)
• StreamRL: Scalable, Heterogeneous, and Elastic RL for LLMs with Disaggregated Stream Generation [55.75008325187133]
  Reinforcement learning (RL) has become the core post-training technique for large language models (LLMs)<n>StreamRL is designed with disaggregation from first principles to address two types of performance bottlenecks.<n> Experiments show that StreamRL improves throughput by up to 2.66x compared to existing state-of-the-art systems.
  arXiv  Detail & Related papers  (2025-04-22T14:19:06Z)
• MM-Eureka: Exploring the Frontiers of Multimodal Reasoning with Rule-based Reinforcement Learning [55.82649731348012]
  We introduce the MMK12 dataset and MM-EUREKA with 7B and 32B parameters.<n>The former is a high-quality multimodal mathematics reasoning dataset featuring diverse knowledge domains with human-verified answers and solution processes.<n>The latter is a multimodal model employing rule-based reinforcement learning utilizing online filtering and two-stage training strategy to enhance training stability.
  arXiv  Detail & Related papers  (2025-03-10T14:23:12Z)
• Action-Quantized Offline Reinforcement Learning for Robotic Skill Learning [68.16998247593209]
  offline reinforcement learning (RL) paradigm provides recipe to convert static behavior datasets into policies that can perform better than the policy that collected the data. In this paper, we propose an adaptive scheme for action quantization. We show that several state-of-the-art offline RL methods such as IQL, CQL, and BRAC improve in performance on benchmarks when combined with our proposed discretization scheme.
  arXiv  Detail & Related papers  (2023-10-18T06:07:10Z)
• SRL: Scaling Distributed Reinforcement Learning to Over Ten Thousand Cores [13.948640763797776]
  We present a novel abstraction on the dataflows of RL training, which unifies diverse RL training applications into a general framework. We develop a scalable, efficient, and distributed RL system called ReaLly scalableRL, which allows efficient and massively parallelized training. SRL is the first in the academic community to perform RL experiments at a large scale with over 15k CPU cores.
  arXiv  Detail & Related papers  (2023-06-29T05:16:25Z)
• RL$^3$: Boosting Meta Reinforcement Learning via RL inside RL$^2$ [12.111848705677142]
  We propose RL$3$, a hybrid approach that incorporates action-values, learned per task via traditional RL, in the inputs to Meta-RL.<n>We show that RL$3$ earns a greater cumulative reward in the long term compared to RL$2$ while drastically reducing meta-training time and generalizes better to out-of-distribution tasks.
  arXiv  Detail & Related papers  (2023-06-28T04:16:16Z)
• LCRL: Certified Policy Synthesis via Logically-Constrained Reinforcement Learning [78.2286146954051]
  LCRL implements model-free Reinforcement Learning (RL) algorithms over unknown Decision Processes (MDPs) We present case studies to demonstrate the applicability, ease of use, scalability, and performance of LCRL.
  arXiv  Detail & Related papers  (2022-09-21T13:21:00Z)
• RL-DARTS: Differentiable Architecture Search for Reinforcement Learning [62.95469460505922]
  We introduce RL-DARTS, one of the first applications of Differentiable Architecture Search (DARTS) in reinforcement learning (RL) By replacing the image encoder with a DARTS supernet, our search method is sample-efficient, requires minimal extra compute resources, and is also compatible with off-policy and on-policy RL algorithms, needing only minor changes in preexisting code. We show that the supernet gradually learns better cells, leading to alternative architectures which can be highly competitive against manually designed policies, but also verify previous design choices for RL policies.
  arXiv  Detail & Related papers  (2021-06-04T03:08:43Z)
• Towards Standardizing Reinforcement Learning Approaches for Stochastic Production Scheduling [77.34726150561087]
  reinforcement learning can be used to solve scheduling problems. Existing studies rely on (sometimes) complex simulations for which the code is unavailable. There is a vast array of RL designs to choose from. standardization of model descriptions - both production setup and RL design - and validation scheme are a prerequisite.
  arXiv  Detail & Related papers  (2021-04-16T16:07:10Z)

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.