Confirming Claims of Superposition and Adversarial Examples in Toy Models

A replication of core claims about superposition in toy models, with an interactive animation of a PGD adversarial attack unfolding in the model’s latent space.

Reading Su Zizhan

人在美国 读苏东坡

The Homeless

我在每个人身上看到披风。

Time Poverty Is the Greatest Poverty

时间贫困是最大的贫困

My First Rain Night in the US

来美国的第一个雨夜

Grade Chaser

做题家如何理解故乡

What I Learnt in Seven Years at Tsinghua

清华七年，我学会了什么

Not Sure if I Miss It Here

不确定是否怀念这里

Relearning C++

Some notes about C++ and Leetcode.

Hello World

This is my first post.