Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2505.20347 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918303076712448 |
|---|---|
| author | Fang, Wenkai Liu, Shunyu Zhou, Yang Zhang, Kongcheng Zheng, Tongya Chen, Kaixuan Song, Mingli Tao, Dacheng |
| author_facet | Fang, Wenkai Liu, Shunyu Zhou, Yang Zhang, Kongcheng Zheng, Tongya Chen, Kaixuan Song, Mingli Tao, Dacheng |
| contents | Recent advances have demonstrated the effectiveness of Reinforcement Learning (RL) in improving the reasoning capabilities of Large Language Models (LLMs). However, existing works inevitably rely on high-quality instructions and verifiable rewards for effective training, both of which are often difficult to obtain in specialized domains. In this paper, we propose Self-play Reinforcement Learning (SeRL) to bootstrap LLM training with limited initial data. Specifically, SeRL comprises two complementary modules: self-instruction and self-rewarding. The former module generates additional instructions based on the available data at each training step, employing robust online filtering strategies to ensure instruction quality, diversity, and difficulty. The latter module introduces a simple yet effective majority-voting mechanism to estimate response rewards for additional instructions, eliminating the need for external annotations. Finally, SeRL performs conventional RL based on the generated data, facilitating iterative self-play learning. Extensive experiments on various reasoning benchmarks and across different LLM backbones demonstrate that the proposed SeRL yields results superior to its counterparts and achieves performance on par with those obtained by high-quality data with verifiable rewards. Our code is available at https://github.com/wantbook-book/SeRL. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2505_20347 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | SeRL: Self-Play Reinforcement Learning for Large Language Models with Limited Data Fang, Wenkai Liu, Shunyu Zhou, Yang Zhang, Kongcheng Zheng, Tongya Chen, Kaixuan Song, Mingli Tao, Dacheng Computation and Language Artificial Intelligence Recent advances have demonstrated the effectiveness of Reinforcement Learning (RL) in improving the reasoning capabilities of Large Language Models (LLMs). However, existing works inevitably rely on high-quality instructions and verifiable rewards for effective training, both of which are often difficult to obtain in specialized domains. In this paper, we propose Self-play Reinforcement Learning (SeRL) to bootstrap LLM training with limited initial data. Specifically, SeRL comprises two complementary modules: self-instruction and self-rewarding. The former module generates additional instructions based on the available data at each training step, employing robust online filtering strategies to ensure instruction quality, diversity, and difficulty. The latter module introduces a simple yet effective majority-voting mechanism to estimate response rewards for additional instructions, eliminating the need for external annotations. Finally, SeRL performs conventional RL based on the generated data, facilitating iterative self-play learning. Extensive experiments on various reasoning benchmarks and across different LLM backbones demonstrate that the proposed SeRL yields results superior to its counterparts and achieves performance on par with those obtained by high-quality data with verifiable rewards. Our code is available at https://github.com/wantbook-book/SeRL. |
| title | SeRL: Self-Play Reinforcement Learning for Large Language Models with Limited Data |
| topic | Computation and Language Artificial Intelligence |
| url | https://arxiv.org/abs/2505.20347 |