Paper Detail
Lei Bai, Jiaqi Cao, Chiyu Chen, Guanzhou Chen, Kai Chen, Guangran Cheng, Erfei Cui, Xuanlang Dai, Shengyuan Ding, Shangheng Du, Yanhui Duan, Yue Fan, Youqing Fang, Quan Gan, Yuanyuan Gao, Jiaye Ge, Lixin Gu, Yuzhe Gu, Qipeng Guo, Junjun He, Xin Hong, Ming Hu, Zhouqi Hua, Haian Huang, Junhao Huang, Zixian Huang, Minxi Jin, Lingkai Kong, Alexander Lam, Zehao Li, Zonglin Li, Tianhao Liang, Dahua Lin, Junyao Lin, Tianyang Lin, Zhouhan Lin, Jiangning Liu, Jin Liu, Kuikun Liu, Wenran Liu, Yifei Liu, Yuhong Liu, Zhoumianze Liu, Ziyan Liu, Ziyu Liu, Haijun Lv, Han Lv, Chengqi Lyu, Le Ma, Ningsheng Ma, Zerun Ma, Haoyang Peng, Runyu Peng, Jifei Shan, Zixin Shang, Kou Shi, Xiang Shi, Qisheng Su, Xuerui Su, Hao Sun, Xiao Sun, Yanan Sun, Yu Sun, Huanze Tang, Yinghao Tang, Wenhui Tian, Zhongbo Tian, Bingli Wang, Haomin Wang, Jiarui Wang, Jingzhi Wang, Rui Wang, Xiquan Wang, Yi Wang, Zhecan Wang, Ziyi Wang, Zun Wang, Rubin Wei, Lianyi Wu, Wen Wu, Yue Wu, Yuhan Wu, Zhenyu Wu, Zijian Wu, Shuhao Xing, Jun Xu, Xingle Xu, Xuenan Xu, Xiangchao Yan, Ziang Yan, Bowen Yang, Danni Yang, Lin Yang, Zhiqi Yang, Qian Yao, Haochen Ye, Peng Ye, Jinhui Yin, Jiashuo Yu, Dingbo Yuan, Fei Yuan, Yuhang Zang, Bo Zhang, Chao Zhang, Chen Zhang, Hongjie Zhang, Junming Zhang, Wenlong Zhang, Wenwei Zhang, Yiming Zhang, Zhuo Zhang, Ziyang Zhang, Haiteng Zhao, Penghao Zhao, Yibo Zhao, Zhonghan Zhao, Zhihang Zhong, Bowen Zhou, Peiheng Zhou, Xin Zhou, Xinyu Zhou, Yunhua Zhou, Dongsheng Zhu, Yicheng Zou
Scientific discovery increasingly requires AI systems that can reason over scientific evidence of heterogeneous modalities, interact with scientific tools and environments, and sustain progress across long task horizons. We present Intern-S2-Preview, a series of scientific agentic foundation models designed to support multimodal scientific understanding, reasoning, generation, and long-horizon tasks. The training pipeline begins with scientific multimodal pre-training over rendered scientific documents, interleaved image-text data, and diverse scientific corpora. Starting from the pretrained checkpoint, we apply a unified post-training pipeline consisting of supervised fine-tuning, scalable multi-task reinforcement learning (RL), black- and white-box agentic RL, and on-policy distillation. This pipeline is supported by practical techniques that improve rollout and training stability and efficiency, including partial rollout with off-policy correction, adaptive length regularization, online speculative decoding, robust multi-task optimization, and trace-aware experience assembly for agentic tasks. At the architecture level, Intern-S2-Preview-397B extends time series modelling from efficient long-sequence understanding to numerical forecasting, while Memory Decoder is studied as a separate memory-augmented path for rapid scientific specialization without modifying the frozen 397B backbone. Evaluations across scientific, multimodal, agentic, and general-purpose benchmarks show that Intern-S2-Preview-397B achieves competitive or leading results in multiple settings. The time series modules improve scientific signal understanding and forecasting on SciTS, while the separate Intern-MemDec-4B extension improves the Biology-Instructions average score from 56.92 to 60.32 without modifying the frozen 397B backbone.
No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.
No ranking explanation is available yet.
No tags.
@article{bai2026intern,
title = {Intern-S2-Preview: Scientific Agentic Foundation Model},
author = {Lei Bai and Jiaqi Cao and Chiyu Chen and Guanzhou Chen and Kai Chen and Guangran Cheng and Erfei Cui and Xuanlang Dai and Shengyuan Ding and Shangheng Du and Yanhui Duan and Yue Fan and Youqing Fang and Quan Gan and Yuanyuan Gao and Jiaye Ge and Lixin Gu and Yuzhe Gu and Qipeng Guo and Junjun He and Xin Hong and Ming Hu and Zhouqi Hua and Haian Huang and Junhao Huang and Zixian Huang and Minxi Jin and Lingkai Kong and Alexander Lam and Zehao Li and Zonglin Li and Tianhao Liang and Dahua Lin and Junyao Lin and Tianyang Lin and Zhouhan Lin and Jiangning Liu and Jin Liu and Kuikun Liu and Wenran Liu and Yifei Liu and Yuhong Liu and Zhoumianze Liu and Ziyan Liu and Ziyu Liu and Haijun Lv and Han Lv and Chengqi Lyu and Le Ma and Ningsheng Ma and Zerun Ma and Haoyang Peng and Runyu Peng and Jifei Shan and Zixin Shang and Kou Shi and Xiang Shi and Qisheng Su and Xuerui Su and Hao Sun and Xiao Sun and Yanan Sun and Yu Sun and Huanze Tang and Yinghao Tang and Wenhui Tian and Zhongbo Tian and Bingli Wang and Haomin Wang and Jiarui Wang and Jingzhi Wang and Rui Wang and Xiquan Wang and Yi Wang and Zhecan Wang and Ziyi Wang and Zun Wang and Rubin Wei and Lianyi Wu and Wen Wu and Yue Wu and Yuhan Wu and Zhenyu Wu and Zijian Wu and Shuhao Xing and Jun Xu and Xingle Xu and Xuenan Xu and Xiangchao Yan and Ziang Yan and Bowen Yang and Danni Yang and Lin Yang and Zhiqi Yang and Qian Yao and Haochen Ye and Peng Ye and Jinhui Yin and Jiashuo Yu and Dingbo Yuan and Fei Yuan and Yuhang Zang and Bo Zhang and Chao Zhang and Chen Zhang and Hongjie Zhang and Junming Zhang and Wenlong Zhang and Wenwei Zhang and Yiming Zhang and Zhuo Zhang and Ziyang Zhang and Haiteng Zhao and Penghao Zhao and Yibo Zhao and Zhonghan Zhao and Zhihang Zhong and Bowen Zhou and Peiheng Zhou and Xin Zhou and Xinyu Zhou and Yunhua Zhou and Dongsheng Zhu and Yicheng Zou},
year = {2026},
abstract = {Scientific discovery increasingly requires AI systems that can reason over scientific evidence of heterogeneous modalities, interact with scientific tools and environments, and sustain progress across long task horizons. We present Intern-S2-Preview, a series of scientific agentic foundation models designed to support multimodal scientific understanding, reasoning, generation, and long-horizon tasks. The training pipeline begins with scientific multimodal pre-training over rendered scientific do},
url = {https://arxiv.org/abs/2608.13505},
keywords = {cs.LG, cs.CL, cs.CV},
eprint = {2608.13505},
archiveprefix = {arXiv},
}
{}