跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.21072 · v1

Sage S2.1 Flash Technical Report

Kun CaiHaodong ChenGuangran ChengQingxiu ChengXu DaiGuanting DongQiming GeLixin GuTao GuiHonglin GuoYuyang HuZixian HuangJiajie JinAlexander LamJiaqiang LiYining LiJiahang LinYanjiang LiuXinyu LuHaijun LvNingsheng MaZerun MaMinlong PengJianmin QianHuayu ShaJunlin ShangShuo ShenQisheng SuJiajun SunYanan SunZhichao SunRan TaoZhongbo TianGuoqiang WangHui WangRui WangYiding WangZhecan WangHao XiangXinchen XieShuhao XingXiaoyu XingWanghan XuXinyu YangYajie YangDingbo YuanYuhang ZangHongjie ZhangChengfeng ZhaoHaoran ZhaoPenghao ZhaoRuojun ZhouDongsheng ZhuYicheng ZouQiye CaiYifei CaoXinmeng CheHongji ChenJiabei ChenJiadong ChenJiahao ChenJiayi ChenXinjie ChenYujia ChenXin DengYi DongYichen DongWenjie FuJianing GengChenyang GuDing HanHan HanFeiyang HaoHaotan HeQiushi HeXufeng HeXin HongJie HouBinze HuYongkang HuZijian HuEnfu HuangJunhao HuangChanghao JiangHuicheng JiangJiazhen JiangShufan JiangWenqing JingJiahao KuangBowen LaiBo LiQi LiQilong LiZhuoqun LiTianhao LiangJiaxiang LiuJunzhang LiuShaofan LiuShuainan LiuTong LiuYi LiuZiyu LiuZhonghang LuJianwen LuoXuan LuoYanyi LuoHuijie LvMingyu MaHoucheng MinHongjun MoSonghao NiChengjun PanQiyuan PengXiaoxuan PengWanying RenRunhui RuanJifei ShanZixin ShangBing ShaoZhuohui ShengJiayang ShiQingyun ShiYang ShuAierpanjiang SimayiSirui SongYuxiao SongJiahao SunZhe SunWenzhe TanGuoxiang WangHanchen WangJie WangPengbo WangQing WangRui WangWeihang WangXiaotao WangYuhui WangZhekai WangZihan WangZile WangJinke WuLiwei WuXuansheng WuYida WuYihao WuYinqi WuYue WuYuhan WuYuwa WuZhiheng XiYing XiaoZhide XieCaijun XuXingle XuYichen XuYongfeng XuXiaolei YangZhixiong YangJiacheng YaoDingcheng YiShihong YiJinhui YinPeizhou YuShan YuHao YuanJunjie YuanXiongyao ZhaBo ZhangCaixian ZhangKaicheng ZhangPengxiang ZhangQiuyinzhe ZhangWei ZhangXiaoqian ZhangXinsen ZhangJiyuan ZhaoShuyu ZhaoPujun ZhengZheyu ZhengXiaoxue ZhongPeiheng ZhouXiaohao ZhouXinyu ZhouFangyi ZhuGuanru ZhuYulun ZhuJiabao ZhuangPengfei CaoZhicheng DouXianpei HanBen HeGuoxiu HeXuanjing HuangTao JiYu-Gang JiangDahua LinHongyu LinKang LiuYaojie LuLe SunJi-Rong WenJun ZhaoYutao ZhuQi ZhangBowen Zhou
首次提交 2026-11-16 21:23
版本更新 2026-11-16 21:23
雷达观测 2026-11-17 17:11
arXiv ↗PDF ↗

摘要

We introduce Sage S1.5 Flash, a 320-billion-parameter general-purpose mixture-of-experts model designed to reason over scientific evidence and act in interactive environments. Sage Sci-Perception uses dedicated encoders for molecular graphs and biological sequences and represents spectra as ordered patches of measurements. Modality-specific projectors map these representations into the input space of a shared backbone, which processes them alongside text and images. To develop the model's agent capabilities, we construct verifiable environments for tool use, search, and coding, as well as specialized workflows involving machine learning experiments and scientific software. Task completion is checked against execution results, changes in environment state, or submitted artifacts. We evaluate Sage S1.5 Flash on 33 benchmarks across seven domains. In the primary comparison against five baselines, it ranks first on OfficeQA Pro (72.2), MLE-bench Lite (90.3 HumanRank), Harvey LAB, and SpreadsheetBench~2, and second on OSWorld-Verified (85.2), WebArena-Verified (67.6), and DeepResearch Bench~II (54.2). These results show particular strengths in using documentary evidence, producing verifiable workbooks, and carrying out machine learning tasks, alongside competitive performance in desktop and browser workflows that require the model to act on visual feedback.

作者备注:website: https://internlm.github.io/Sage-S2.1-Flash

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-16 21:23 · 最新观测版本

Sage S2.1 Flash Technical Report

We introduce Sage S1.5 Flash, a 320-billion-parameter general-purpose mixture-of-experts model designed to reason over scientific evidence and act in interactive environments. Sage Sci-Perception uses dedicated encoders for molecular graphs and biological sequences and represents spectra as ordered patches of measurements. Modality-specific projectors map these representations into the input space of a shared backbone, which processes them alongside text and images. To develop the model's agent capabilities, we construct verifiable environments for tool use, search, and coding, as well as specialized workflows involving machine learning experiments and scientific software. Task completion is checked against execution results, changes in environment state, or submitted artifacts. We evaluate Sage S1.5 Flash on 33 benchmarks across seven domains. In the primary comparison against five baselines, it ranks first on OfficeQA Pro (72.2), MLE-bench Lite (90.3 HumanRank), Harvey LAB, and SpreadsheetBench~2, and second on OSWorld-Verified (85.2), WebArena-Verified (67.6), and DeepResearch Bench~II (54.2). These results show particular strengths in using documentary evidence, producing verifiable workbooks, and carrying out machine learning tasks, alongside competitive performance in desktop and browser workflows that require the model to act on visual feedback.

查看此版本解读 →查看此版本原文 ↗