paper-with-me

Papers

Accelerating RL Post-Training Rollouts via System-Integrated Speculative Decoding

2026-04-29 · Hayate Iso, Tiyasa Mitra, Sudipta Mondal, Rasoul Shafipour, Venmugil Elango, Terry Kong, Yuki Huang, Seonjin Na, Izzy Putterman, Benjamin Chislett, Maor Ashkenazi, Joseph Guman, Gerald Shen, Tugrul Konuk, Ashwath Aithal, Ritika Borkar, Ran Zilberstein, Bita Rouhani arxiv

RL post-training of frontier language models is increasingly bottlenecked by autoregressive rollout generation, making rollout acceleration a central systems challenge. Many existing efficiency methods improve throughput by changing the rollout or optimization regime, for example, through off-policy execution, replay, or lower-precision generation. We study speculative decoding as a lossless acceleration primitive for RL rollouts that preserves the target model's output distribution. We implement speculative decoding in NeMo-RL with a vLLM backend, supporting both synchronous and asynchronous pipelines and enabling speculation during RL rollouts. This benefit is realizable across speculation mechanisms, such as pretrained MTP heads, small external draft models or even techniques such as Eagle3, which are traditionally applied after RL phase. This yields a deployment path for state-of-the-art speculative decoding inside RL training. In a reasoning post-training workload at 8B scale under synchronous RL, speculative decoding improves rollout throughput by 1.8x. Using a high-fidelity performance simulator, we project that combining speculative decoding with asynchronous RL yields up to 2.5x end-to-end training speedup at 235B scale.

📄 PDF Abstract BibTeX arXiv:2604.26779

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

SRT: Accelerating Reinforcement Learning via Speculative Rollout with Tree-Structured Cache

2026-01-14 · Chi-Chih Chang, Siqi Zhu, Zhichen Zeng, Haibin Lin 외 arxiv

We present Speculative Rollout with Tree-Structured Cache (SRT), a simple, model-free approach to accelerate on-policy reinforcement learning (RL) for language models without sacrificing distributional correctness. SRT e…

Reinforcement Learning

EfficientRollout: System-Aware Self-Speculative Decoding for RL Rollouts

2026-06-17 · Minseo Kim, Minjae Lee, Seunghyuk Oh, Kevin Galim 외 arxiv

Reinforcement learning (RL) has become a representative post-training paradigm for LLMs, enabling strong reasoning and agentic capabilities. However, rollout generation remains a dominant latency bottleneck because autor…

Reinforcement Learning

Learning from the Right Rollouts: Data Attribution for PPO-based LLM Post-Training

2026-04-02 · Dong Shu, Denghui Zhang, Jessica Hullman arxiv

Traditional RL algorithms like Proximal Policy Optimization (PPO) typically train on the entire rollout buffer, operating under the assumption that all generated episodes provide a beneficial optimization signal. However…

FP4 Explore, BF16 Train: Diffusion Reinforcement Learning via Efficient Rollout Scaling

2026-04-08 · Yitong Li, Junsong Chen, Shuchen Xue, Pengcuo Zeren 외 arxiv

Reinforcement-Learning-based post-training has recently emerged as a promising paradigm for aligning text-to-image diffusion models with human preferences. In recent studies, increasing the rollout group size yields pron…

Reinforcement Learning

EchoRL: Reinforcement Learning via Rollout Echoing

2026-05-29 · Jinhe Bi, Aniri, Minglai Yang, Xingcheng Zhou 외 arxiv

Reinforcement Learning with Verifiable Rewards is an effective route for post-training to strengthen the reasoning capability of large language models. However, as training proceeds, the learning signal can collapse thus…

Reinforcement Learning