The Great Library

todo list

too lazy to edit :)

Agent S2 uses expert model for grounding and replan when subgoal finishes.

This is an independent score computing method as a supplement to OSWorld. In this work, the authors design a brand new metric to evaluate agent's work efficiency.

This document records the main procedure of AI development, from neural network to CNN to RNN to Transformer to LLM to agent...

Too lazy to describe.

This article concludes the workflow of building a Hexo-based blog.

A practical guide to my personal knowledge graph: notes, edges, evidence, queries, and how it connects to this blog.

Theory of Machine Learning & Deep Learning & Reinforcement Learning

Describe the mechanism of Agent architecture.