Preprint · 2025 · Open access
Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models
arXiv (Cornell University) · Published 1 August 2025 · DOI 10.48550/arxiv.2508.00410
In brief
“Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models” is a 2025 preprint by Zizhuo Zhang and 8 co-authors, published in arXiv (Cornell University). OpenAlex counts 33 citations as of 15 November 2026. It is open access under a CC BY licence: the full text is free to read at arXiv (Cornell University).
- Citations
- 33
- Most cited in
- 2026 (25)
- References
- 45
- Authors
- 9
Abstract
While reinforcement learning with verifiable rewards (RLVR) is effective to improve the reasoning ability of large language models (LLMs), its reliance on human-annotated labels leads to the scaling up dilemma, especially for complex tasks. Recent self-rewarding methods investigate a label-free alternative to unlock the reasoning capabilities of LLMs, yet they frequently encounter the non-negligible training collapse issue, as the single-view supervision signal easily forms the self-consistent illusion, yielding the reward hacking. Inspired by the success of self-supervised learning, we propose \textit{Co-rewarding}, a novel self-supervised RL framework that improves training stability by seeking complementary supervision from another views. Specifically, we instantiate Co-rewarding in two ways: (1) \textit{Co-rewarding-I} is a data-side instantiation that derives reward signals from contrastive agreement across semantically analogous questions; and (2) \textit{Co-rewarding-II} is a model-side instantiation that maintains a slowly-updated reference teacher with pseudo labels to realize self-distillation. Intuitively, such instantiations introduce different levels of discrepancy to increase the difficulty of training collapse on trivial reasoning solutions. We also explore their orthogonally combined version to further boost the performance. Empirically, Co-rewarding exhibits stable training across various setups, and outperforms other self-rewarding baselines by $+3.31\%$ improvements on average on multiple mathematical reasoning benchmarks, especially by $+7.49\%$ on Llama-3.2-3B-Instruct. Notably, Co-rewarding reaches or even surpasses RLVR with ground-truth (GT) label in several cases, such as a Pass@$1$ of $94.01\%$ on GSM8K with Qwen3-8B-Base remarkably higher than GT. Our code is released at https://github.com/tmlr-group/Co-rewarding.
Cite this paper
Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.
APA 7
Zhang, Z., Zhu, J., Xinmu, G., Zhao, Z., Zhou, Z., Li, X., Xiao, F., Yao, J., & Han, B. (2025). Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2508.00410
MLA 9
Zhang, Zizhuo, et al. “Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models.” arXiv (Cornell University), 2025. https://doi.org/10.48550/arxiv.2508.00410.
Chicago
Zhang, Zizhuo, Jianing Zhu, Ge, Xinmu, Zihua Zhao, Zhanke Zhou, Xuan Li, Feng Xiao, Jiangchao Yao, and Bo Han. 2025. “Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models.” arXiv (Cornell University). https://doi.org/10.48550/arxiv.2508.00410.
Harvard
Zhang, Z. et al. (2025) ‘Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models’, arXiv (Cornell University). doi:10.48550/arxiv.2508.00410.
IEEE
Z. Zhang et al., “Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models,” arXiv (Cornell University), 2025, doi: 10.48550/arxiv.2508.00410.
BibTeX
@misc{zhang2025rewarding,
title = {{Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models}},
author = {Zhang, Zizhuo and Zhu, Jianing and Xinmu, Ge, and Zhao, Zihua and Zhou, Zhanke and Li, Xuan and Xiao, Feng and Yao, Jiangchao and Han, Bo},
howpublished = {arXiv (Cornell University)},
year = {2025},
doi = {10.48550/arxiv.2508.00410},
url = {https://doi.org/10.48550/arxiv.2508.00410}
}
Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)
Topic and keywords
Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- large language models
- reward hacking
- pseudo-labeling
- self-distillation
- teacher-student learning
- mathematical reasoning
Work with this paper
Most-cited papers in Computer Science
- ImageNet classification with deep convolutional neural networksAlex Krizhevsky, Ilya Sutskever, Geoffrey E. Hinton · Communications of the ACM · 2017 · 114,726 citations
- Long Short-Term MemorySepp Hochreiter, Jürgen Schmidhuber · Neural Computation · 1997 · 109,097 citations
- A Mathematical Theory of CommunicationClaude E. Shannon · Bell System Technical Journal · 1948 · 91,736 citations
- Deep learningYann LeCun, Yoshua Bengio, Geoffrey E. Hinton · Nature · 2015 · 90,226 citations
- Faster R-CNN: Towards Real-Time Object Detection with Region Proposal NetworksShaoqing Ren, Kaiming He, Ross Girshick and 1 more · IEEE Transactions on Pattern Analysis and Machine Intelligence · 2016 · 74,143 citations
Questions about this paper
How do I cite this paper in APA style?
Zhang, Z., Zhu, J., Xinmu, G., Zhao, Z., Zhou, Z., Li, X., Xiao, F., Yao, J., & Han, B. (2025). Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models. arXiv (Cornell University). https://doi.org/10.48550/arxiv.2508.00410
How many times has this paper been cited?
33 times, according to OpenAlex on 15 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.
Is this paper free to read?
Yes. It is open access: read the free full text at arXiv (Cornell University).
Who wrote this paper?
Zizhuo Zhang, Jianing Zhu, Ge, Xinmu, Zihua Zhao, Zhanke Zhou, Xuan Li, Feng Xiao, Jiangchao Yao, Bo Han.
What is the DOI of this paper?
10.48550/arxiv.2508.00410. The permanent link is https://doi.org/10.48550/arxiv.2508.00410.
Bibliographic data from OpenAlex (CC0), last checked 15 November 2026. ResearchGuru does not host this paper; the links go to the publisher or the repository that does.