paper-with-me

Papers

Sparrow: Sparse Rollout for Stable and Efficient Long-context RL of Large Language Models

2026-06-07 · Yang Zhou, Ranajoy Sadhukhan, Zhaofeng Sun, Zhuoming Chen, Souvik Kundu, Saket Dingliwal, Sai Muralidhar Jayanthi, Aram Galstyan, Haizhong Zheng, Beidi Chen arxiv

Despite being powerful, reinforcement learning with verifiable rewards (RLVR) induces extremely long COT, making it computationally expensive. Since RLVR per-step cost is dominated by long-context rollout generation, sparse attention offers a promising way to accelerate dense rollout. However, sparse rollouts require a delicate stability-efficiency tradeoff: overly aggressive sparsity causes collapse, while overly lenient sparsity gives insufficient speedup. In this work, we study this tradeoff through sparse-to-dense actor-policy mismatch. We first observe that sparse rollout collapse is not driven by uniform degradation across tokens: most sparse tokens align perfectly with dense even under aggressive sparsity. Motivated by this, we hypothesize that sparse rollout training remains stable if the lower tail of per-token actor-policy mismatch stays above a critical threshold throughout the trajectory. We introduce a dynamic sparsity schedule that keeps this tail statistic constant during generation and validate our hypothesis. Across Qwen3 thinking-family models, keeping the tail mismatch statistic near a consistent threshold generally enables stable training. We then use a cost model to find the sparsity schedule for maximum speedup under this mismatch threshold, achieving 2.2x, 2.4x, and 2.0x rollout speedups when training Qwen3-1.7B, Qwen3-4B, and Qwen3-8B. Empirically, we show the thresholds generalize to a larger model (Qwen3-14B) and another RL domain (coding). Finally, our analysis naturally motivates DistillSparse: lightweight LoRA-based distillation on sparse rollout lets more aggressive sparsity reach the same sparse-to-dense mismatch threshold, yielding higher speedup.

📄 PDF Abstract BibTeX arXiv:2606.08446

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Similar Papers 제목 키워드 기반

SPARROW: Learning Spatial Precision and Temporal Referential Consistency in Pixel-Grounded Video MLLMs

2026-03-12 · Mohamad Alansari, Naufal Suryanto, Divya Velayudhan, Sajid Javed 외 arxiv

Multimodal large language models (MLLMs) have advanced from image-level reasoning to pixel-level grounding, but extending these capabilities to videos remains challenging as models must achieve spatial precision and temp…

Visual Grounding

Sparse-RL: Breaking the Memory Wall in LLM Reinforcement Learning via Stable Sparse Rollouts

2026-01-15 · Sijia Luo, Xiaokang Zhang, Yuxuan Hu, Bohan Zhang 외 arxiv

Reinforcement Learning (RL) has become essential for eliciting complex reasoning capabilities in Large Language Models (LLMs). However, the substantial memory overhead of storing Key-Value (KV) caches during long-horizon…

Reinforcement Learning

A Recipe for Long-Context Reasoning in Large Language Models via On-Policy Optimization and Distillation

2026-05-12 · Miguel Moura Ramos, Duarte M. Alves, André F. T. Martins arxiv

Existing approaches to post-train models for long-context tasks face complementary limitations: (i) supervised fine-tuning (SFT) provides stable supervision but suffers from exposure bias; (ii) reinforcement learning met…

Reinforcement Learning

DemoEvolve: Overcoming Sparse Feedback in Agentic Harness Evolution with Demonstrations

2026-05-23 · Lirong Che, Yuzhe yang, Peiwen lin, Chuang wang 외 arxiv

Agent harness evolution improves frozen language-model agents by modifying the executable structures around them. We study this paradigm as a form of sample-efficient fast adaptation: instead of updating model weights, a…

$V_{0.5}$: Generalist Value Model as a Prior for Sparse RL Rollouts

2026-03-11 · Yi-Kai Zhang, Yueqing Sun, Hongyan Hao, Qi Gu 외 arxiv

In Reinforcement Learning with Verifiable Rewards (RLVR), constructing a robust advantage baseline is critical for policy gradients, effectively guiding the policy model to reinforce desired behaviors. Recent research ha…

Computational EfficiencyReinforcement LearningMathematical Reasoning