Skip to content

Commit fe01f63

Browse files
author
unknown
committed
update
1 parent d9eee16 commit fe01f63

File tree

1 file changed

+1
-2
lines changed

1 file changed

+1
-2
lines changed

news_small.md

Lines changed: 1 addition & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,8 @@
11
# 从 AGI 到 SGI:Gemini‑3‑Pro 拿下 SOTA,却仍然显著不及格
22

3-
- 项目主页 Page:https://internscience.github.io/SGI-Page/
3+
- 项目主页 Page:https://InternScience.github.io/SGI-Page/
44
- 代码 Code:https://github.com/InternScience/SGI-Bench
55
- 数据 Data:https://huggingface.co/collections/InternScience/sgi-bench
6-
- 团队 Team:https://discovery.intern-ai.org.cn/sciprismax
76

87
近年来,大模型在多学科知识理解、数学推理、编程等任务上频频刷榜,但AI的“科学通用能力”仍没有统一刻度:它是多学科、长链路、跨模态且要求严谨可验证的。许多现有 benchmark 只覆盖碎片能力(如学科问答、单步工具操作),难以映射到真实科研中的循环与自纠错过程。为此,我们通过引入实践探究模型(PIM),将科学探究过程拆解为四个循环的阶段,并与AI的能力维度进行了对应:
98

0 commit comments

Comments
 (0)