🔬
Intern Discovery-EarthLink

🔬

Platform Address:https://earthlink.intern-ai.org.cn/

Problem Feedback: https://aicarrier.feishu.cn/share/base/form/shrcnDbTGK37ZrjkAvZJqjyZH8d

Article Link: https://arxiv.org/abs/2507.17311

👋 Welcome to Intern Discovery-EarthLink 👋

1.
Community Invitation and Demand Feedback​

😍

Thank you all for your usage and attention. We will uniformly open a batch of user access rights every Wednesday. Currently, the number of our users has exceeded95. We welcome you to recommend each other!​

WeChat groups provide us with the latest updates in real time

common.docs_name - LarkCCM_Docs_Menu_Image

33%

If you don't have WeChat, you can contact us via email or github "issue"

lingfenghua@pjlab.org.cn

guozijie@pjlab.org.cn

33%

Functional Version Update:

Earth Scientist Version 0.2.0 has been successfully launched

🌍Core Functional Buttons in This Issue:

a.
A new research plan section has been added, which supports viewing research plan items before generating analysis reports ​

b.
New human-machine interaction design has been added. In semi-automatic mode, it supports users to manually modify the research plan before performing analysis and generating a research analysis report. ​

c.
New user history added, supporting user backtracking​

d.
Added user feedback, help documentation, dataset introduction, and GitHub referral link​

e.
New file display function added, supporting users to view code files and data files generated during the entire intermediate process ​

f.
New scoring functions for analysis reports and research plans have been added, supporting users to evaluate and generate research plans and analysis reports ​

g.
New image-text PDF attachments are added, enabling users to view image-text report attachments and more intuitively view the results of data lake visualization​

2.
Overview​

EarthLink, supported by the "Shusheng" general-purpose and specialized integrated foundational large model, is the first multi-agent system that systematically integrates climate domain data, knowledge, tools, and general large model capabilities. It aims to completely eliminate the barriers to programming and software use, empowering scientists to focus on scientific problems themselves. Through three main modules, namely "human-machine collaborative intelligent planning", "self-evolving experimental platform", and "multi-scenario scientific analysis", the system has realized a "thought-decision" self-evolving workflow starting from natural language instructions, helping researchers, research tools, and research objects to comprehensively enhance their capabilities, evolve collaboratively, and drive scientific research from team single-point exploration towards the Scaling Law stage of scientific discovery.​