Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2604.17312 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866917420296306688 |
|---|---|
| author | Yu, Zhiyin Mou, Yuchen Yan, Juncheng Luo, Junyu Chen, Chunchun Wei, Xing Liu, Yunhui Sun, Hongru Zhang, Yuxing Xu, Jun Bian, Yatao Zhang, Ming Ye, Wei He, Tieke Yang, Jie Zheng, Guanjie Wu, Zhonghai Zhang, Bo Bai, Lei Luo, Xiao |
| author_facet | Yu, Zhiyin Mou, Yuchen Yan, Juncheng Luo, Junyu Chen, Chunchun Wei, Xing Liu, Yunhui Sun, Hongru Zhang, Yuxing Xu, Jun Bian, Yatao Zhang, Ming Ye, Wei He, Tieke Yang, Jie Zheng, Guanjie Wu, Zhonghai Zhang, Bo Bai, Lei Luo, Xiao |
| contents | Reinforcement learning (RL) has emerged as a powerful post-training paradigm for enhancing the reasoning capabilities of large language models (LLMs). However, reinforcement learning for LLMs faces substantial data scarcity challenges, including the limited availability of high-quality external supervision and the constrained volume of model-generated experience. These limitations make data-efficient reinforcement learning a critical research direction. In this survey, we present the first systematic review of reinforcement learning for LLMs under data scarcity. We propose a bottom-up hierarchical framework built around three complementary perspectives: the data-centric perspective, the training-centric perspective, and the framework-centric perspective. We develop a taxonomy of existing methods, summarize representative approaches in each category, and analyze their strengths and limitations. Our taxonomy aims to provide a clear conceptual foundation for understanding the design space of data-efficient RL for LLMs and to guide researchers working in this emerging area. We hope this survey offers a comprehensive roadmap for future research and inspires new directions toward more efficient and scalable reinforcement learning post-training for LLMs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2604_17312 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | A Survey of Reinforcement Learning for Large Language Models under Data Scarcity: Challenges and Solutions Yu, Zhiyin Mou, Yuchen Yan, Juncheng Luo, Junyu Chen, Chunchun Wei, Xing Liu, Yunhui Sun, Hongru Zhang, Yuxing Xu, Jun Bian, Yatao Zhang, Ming Ye, Wei He, Tieke Yang, Jie Zheng, Guanjie Wu, Zhonghai Zhang, Bo Bai, Lei Luo, Xiao Machine Learning Artificial Intelligence Reinforcement learning (RL) has emerged as a powerful post-training paradigm for enhancing the reasoning capabilities of large language models (LLMs). However, reinforcement learning for LLMs faces substantial data scarcity challenges, including the limited availability of high-quality external supervision and the constrained volume of model-generated experience. These limitations make data-efficient reinforcement learning a critical research direction. In this survey, we present the first systematic review of reinforcement learning for LLMs under data scarcity. We propose a bottom-up hierarchical framework built around three complementary perspectives: the data-centric perspective, the training-centric perspective, and the framework-centric perspective. We develop a taxonomy of existing methods, summarize representative approaches in each category, and analyze their strengths and limitations. Our taxonomy aims to provide a clear conceptual foundation for understanding the design space of data-efficient RL for LLMs and to guide researchers working in this emerging area. We hope this survey offers a comprehensive roadmap for future research and inspires new directions toward more efficient and scalable reinforcement learning post-training for LLMs. |
| title | A Survey of Reinforcement Learning for Large Language Models under Data Scarcity: Challenges and Solutions |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2604.17312 |