| title | Intern-S2-Mobius: Foundation Model with Decoupled Knowledge and Reasoning |
|---|---|
| authors | Kai Chen, Jifeng Ding, Ning Ding, Jiaye Ge, Lixin Gu, Yicheng Gu, Qipeng Guo, Ermo Hua, Haian Huang, Haozheng Hou, Jie Hou, Xiangyu Hong, Che Jiang, Minxi Jin, Cheng Liang, Dahua Lin, Dawei Liu, Kuikun Liu, Chengqi Lv, Haijun Lv, Han Lv, Ningsheng Ma, Biqing Qi, Jianmin Qian, Shiya Su, Youbang Sun, Huanze Tang, Zhongbo Tian, Hanjing Wang, Rui Wang, Ting Wang, Yi Wang, Baiting Wu, Jun Xu, Bowen Yang, Hui Wang, Weida Wang, Haochen Ye, Jiashuo Yu, Shan Yu, Xiaoyi Yu, Qirui Zeng, Qi Zhang, Ming Zhang, Wenwei Zhang, Bowen Zhou, Xinyu Zhou |
| date | 2026-08-14 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2608.14290 |
| upvotes | 36 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| abstract | We introduce Mobius-v0, an architecture that comprises a globally shared Memory (FFN) that stores knowledge vectors and multiple Reasoners (Self-Attn) that iteratively achieve compositional reasoning. Using hidden states as cache and carrier, reasoners repeatedly query memory for required knowledge-vectors, while the knowledge is transmitted back to reasoning operators. Through this knowledge-reasoning-separation architecture, Mobius achieves better knowledge compression and reasoning efficiency. Built upon Mobius-v0 architecture: 1) Our 7B model trained-from-scratch achieves similar downstream score as a 7B Transformer baseline with 62.6% of baseline's training data. 2) Our Intern-S2-Mobius, continually-pretrained from Qwen3.5-35B, achieves similar downstream score while delivering nearly 4x end-to-end inference speedup. |
cite
@misc{chen2026intern,
title={Intern-S2-Mobius: Foundation Model with Decoupled Knowledge and Reasoning},
author={Kai Chen and Jifeng Ding and Ning Ding and Jiaye Ge and Lixin Gu and Yicheng Gu and Qipeng Guo and Ermo Hua and Haian Huang and Haozheng Hou and Jie Hou and Xiangyu Hong and Che Jiang and Minxi Jin and Cheng Liang and Dahua Lin and Dawei Liu and Kuikun Liu and Chengqi Lv and Haijun Lv and Han Lv and Ningsheng Ma and Biqing Qi and Jianmin Qian and Shiya Su and Youbang Sun and Huanze Tang and Zhongbo Tian and Hanjing Wang and Rui Wang and Ting Wang and Yi Wang and Baiting Wu and Jun Xu and Bowen Yang and Hui Wang and Weida Wang and Haochen Ye and Jiashuo Yu and Shan Yu and Xiaoyi Yu and Qirui Zeng and Qi Zhang and Ming Zhang and Wenwei Zhang and Bowen Zhou and Xinyu Zhou},
year={2026},
eprint={2608.14290},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2608.14290}
}