Can I read Parallel-R1: Towards Parallel Thinking via Reinforcement Learning on EtoBox?

Parallel-R1: Towards Parallel Thinking via Reinforcement Learning by Tong Zheng, Hongming Zhang, Wenhao Yu, Xiaoyang Wang, Runpeng Dai, Rui Liu, Huiwen Bao, Chengsong Huang, Heng Huang Dong Yu is a nonfiction available to read on EtoBox.

What is Parallel-R1: Towards Parallel Thinking via Reinforcement Learning about?

arXiv:2509.07980v2 [cs.CL] 12 Sep 2025AbstractParallel thinking has emerged as a novel approach for enhancing the reasoning capabilities of large language models (LLMs) by exploring multiple reasoning paths concurrently.However, activating such capabilities through training remains challenging, as existingmethods predominantly rely on supervised fine-tuning (SFT) over synthetic data, whichencourages teacher-forced imitation rather than exploration and generalization. Different from them, we propose Parallel-R1, the first reinforcement learning (RL) frameworkthat enables parallel thinking behaviors for complex real-world reasoning tasks. Ourframework employs a progressive curriculum that explicitly addresses the cold-startproblem in training parallel thinking with RL. We first use SFT on prompt-generatedtrajectories from easier tasks to instill the parallel thinking ability, then transition to RLto explore and generalize this skill on harder problems. Experiments on various mathbenchmarks, including MATH, AMC23, and AIME, show that Parallel-R1 successfully instills parallel thinking, leading to 8.4% accuracy improvements over the sequential thinking model trained directly on cha

Who reads Parallel-R1: Towards Parallel Thinking via Reinforcement Learning?

Author

Tong Zheng, Hongming Zhang, Wenhao Yu, Xiaoyang Wang, Runpeng Dai, Rui Liu, Huiwen Bao, Chengsong Huang, Heng Huang Dong Yu

Published

2025

Language

EN

Category

nonfiction

Subjects

Computer Science, Artificial Intelligence (Ai), Stem