Vibe Research: Training, Safeguarding, and Evaluating Research Agents in a Lab
A lab-level pattern for making research agents useful: train them with shared context, safeguard them with infrastructure, and evaluate them through expert steering.
Hi, this is Shigang. I'm documenting my learning notes on AI agents, scientific research, and other things I find interesting.
More about me →A lab-level pattern for making research agents useful: train them with shared context, safeguard them with infrastructure, and evaluate them through expert steering.
Notes on agents as general problem solvers, cognitive computing, and embodied cognition.
本文将介绍如何将程序需求自动变成容易修改的流程图,使用的工具是Copilot和Markdown,下面是我得到的结果:
本文是基于最小可行产品概念发散出的科研效率idea,简单来说就是:
用 PageRank 分析网站之间的链接结构,并以中国科学院大学及其相关机构为例,探索网络关系、组织体量和影响力。