Skip to content

Source paper

Sage S2.1 Flash Technical Report

Kun Cai, Haodong Chen, Guangran Cheng and 203 more

Kun Cai, Haodong Chen, Guangran Cheng, Qingxiu Cheng, Xu Dai, Guanting Dong, Qiming Ge, Lixin Gu, Tao Gui, Honglin Guo, Yuyang Hu, Zixian Huang, Jiajie Jin, Alexander Lam, Jiaqiang Li, Yining Li, Jiahang Lin, Yanjiang Liu, Xinyu Lu, Haijun Lv, Ningsheng Ma, Zerun Ma, Minlong Peng, Jianmin Qian, Huayu Sha, Junlin Shang, Shuo Shen, Qisheng Su, Jiajun Sun, Yanan Sun, Zhichao Sun, Ran Tao, Zhongbo Tian, Guoqiang Wang, Hui Wang, Rui Wang, Yiding Wang, Zhecan Wang, Hao Xiang, Xinchen Xie, Shuhao Xing, Xiaoyu Xing, Wanghan Xu, Xinyu Yang, Yajie Yang, Dingbo Yuan, Yuhang Zang, Hongjie Zhang, Chengfeng Zhao, Haoran Zhao, Penghao Zhao, Ruojun Zhou, Dongsheng Zhu, Yicheng Zou, Qiye Cai, Yifei Cao, Xinmeng Che, Hongji Chen, Jiabei Chen, Jiadong Chen, Jiahao Chen, Jiayi Chen, Xinjie Chen, Yujia Chen, Xin Deng, Yi Dong, Yichen Dong, Wenjie Fu, Jianing Geng, Chenyang Gu, Ding Han, Han Han, Feiyang Hao, Haotan He, Qiushi He, Xufeng He, Xin Hong, Jie Hou, Binze Hu, Yongkang Hu, Zijian Hu, Enfu Huang, Junhao Huang, Changhao Jiang, Huicheng Jiang, Jiazhen Jiang, Shufan Jiang, Wenqing Jing, Jiahao Kuang, Bowen Lai, Bo Li, Qi Li, Qilong Li, Zhuoqun Li, Tianhao Liang, Jiaxiang Liu, Junzhang Liu, Shaofan Liu, Shuainan Liu, Tong Liu, Yi Liu, Ziyu Liu, Zhonghang Lu, Jianwen Luo, Xuan Luo, Yanyi Luo, Huijie Lv, Mingyu Ma, Houcheng Min, Hongjun Mo, Songhao Ni, Chengjun Pan, Qiyuan Peng, Xiaoxuan Peng, Wanying Ren, Runhui Ruan, Jifei Shan, Zixin Shang, Bing Shao, Zhuohui Sheng, Jiayang Shi, Qingyun Shi, Yang Shu, Aierpanjiang Simayi, Sirui Song, Yuxiao Song, Jiahao Sun, Zhe Sun, Wenzhe Tan, Guoxiang Wang, Hanchen Wang, Jie Wang, Pengbo Wang, Qing Wang, Rui Wang, Weihang Wang, Xiaotao Wang, Yuhui Wang, Zhekai Wang, Zihan Wang, Zile Wang, Jinke Wu, Liwei Wu, Xuansheng Wu, Yida Wu, Yihao Wu, Yinqi Wu, Yue Wu, Yuhan Wu, Yuwa Wu, Zhiheng Xi, Ying Xiao, Zhide Xie, Caijun Xu, Xingle Xu, Yichen Xu, Yongfeng Xu, Xiaolei Yang, Zhixiong Yang, Jiacheng Yao, Dingcheng Yi, Shihong Yi, Jinhui Yin, Peizhou Yu, Shan Yu, Hao Yuan, Junjie Yuan, Xiongyao Zha, Bo Zhang, Caixian Zhang, Kaicheng Zhang, Pengxiang Zhang, Qiuyinzhe Zhang, Wei Zhang, Xiaoqian Zhang, Xinsen Zhang, Jiyuan Zhao, Shuyu Zhao, Pujun Zheng, Zheyu Zheng, Xiaoxue Zhong, Peiheng Zhou, Xiaohao Zhou, Xinyu Zhou, Fangyi Zhu, Guanru Zhu, Yulun Zhu, Jiabao Zhuang, Pengfei Cao, Zhicheng Dou, Xianpei Han, Ben He, Guoxiu He, Xuanjing Huang, Tao Ji, Yu-Gang Jiang, Dahua Lin, Hongyu Lin, Kang Liu, Yaojie Lu, Le Sun, Ji-Rong Wen, Jun Zhao, Yutao Zhu, Qi Zhang, Bowen Zhou

arXiv 2611.21072 · 202689 upvotes on Hugging Face

Research ideas from this paper (6)

Generated by AI. The paper’s authors did not write these.

End-to-End Benchmarking of Native Scientific Perception Within Interactive Agentic Workflows

Build a benchmark suite pairing raw molecular, sequence, and spectral inputs with multi-step software tools to evaluate whether multimodal agents actually use scientific perception during discovery.

Sage S2.1 Flash Technical Report
🟠Medium🔬Significant ResearchMultiple GPUsNo close match found#AI for science#Agents

Parameter-Efficient Co-Adaptation of Scientific Encoders and MoE Backbone Routing

Co-adapt the top layers of frozen scientific encoders and language routing weights alongside modality projectors using outcome-based reinforcement learning to improve complex scientific reasoning.

Sage S2.1 Flash Technical Report
🔴Hard💡Moderate InnovationLarge clusterNo close match found#Multimodal#AI for science

Subgraph-Decomposed Curriculum RL for Hard Repository-Level Software Engineering

Apply curriculum-based reinforcement learning over decomposed dependency subgraphs of failed repository-generation tasks so coding agents learn multi-file software construction instead of discarding unsolved problems.

Sage S2.1 Flash Technical Report
🔴Hard🔬Significant ResearchLarge clusterNo close match found#Code#Agents

Artifact-to-Claim Cross-Verification and Process Consistency Rewards for Agentic Deliverables

Add a schema-aware final cross-verification stage and consistency rewards to multi-step agent training so reported summaries match detailed execution logs and artifacts.

Sage S2.1 Flash Technical Report
🟡Fairly easy💡Moderate InnovationMultiple GPUsRelated work found#Agents#Data & evaluation

Hermetic Harness-Agent Environment Isolation Against Dependency-Mediated Answer Leakage

Isolate evaluation harness packages from agent-accessible execution environments with runtime taint tracking to prevent agents from leaking answers through shared software dependencies.

Sage S2.1 Flash Technical Report
🟡Fairly easy💡Moderate InnovationSingle GPU or lessNo close match found#Safety & alignment#Agents

Uncertainty-Triggered Visual Frame Retention for Long-Horizon GUI and Hybrid Agents

Replace fixed-interval screenshot-to-text folding with uncertainty-triggered visual frame retention so long-horizon desktop and browser agents preserve fine-grained interface details without overflowing context limits.

Sage S2.1 Flash Technical Report
🟠Medium💡Moderate InnovationLarge clusterRelated work found#Vision#Agents

AI summary

Sage S2.1 Flash is a 320-billion-parameter mixture-of-experts (MoE) model designed to integrate native scientific perception with general-purpose multimodal reasoning and long-horizon agentic execution. It combines dedicated scientific encoders and modality-specific projectors (Sage Sci-Perception) with post-training across large-scale verifiable environments spanning tool use, search, coding, ML engineering, workspace deliverables, and computer use, achieving top-two rankings on many of 33 evaluated benchmarks.

How it works (AI summary)

The model uses a 320B MoE backbone with hybrid sparse and linear attention that processes text, images, and continuous scientific embeddings mapped from frozen domain encoders (e.g., MoleculeSTM, ESM-2, DNABERT-2, RNA-FM) and numerical spectral patches via trainable projectors with optional context, side, and readout branches. Post-training relies on XTuner across verifiable environments synthesized via tool-composition graphs (ToolSpace), forward/reverse search synthesis, containerized coding/ML/cybersecurity sandboxes, and staged visual CUA curricula, followed by Group Relative Policy Optimization (GRPO) on scientific tasks using verifiable outcome rewards.