The Great Library

todo list

too lazy to edit :)

Agent S2 uses an expert model for grounding and replans once each subgoal is completed.

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

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.

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