AI Safety

Few-Shot Awareness in Reasoning Models

less than 1 minute read

Instruction following inside CoT for better Model Organisms

less than 1 minute read

Towards Mitigating Information Leakage When Evaluating Safety Monitors

less than 1 minute read

AISC: Probing for Deception Detection

less than 1 minute read

Interpretability Hackathon

less than 1 minute read

AI Policy Hackathon

less than 1 minute read

Mechanistic Exploration Gemma 2 List Generation

18 minute read

Abstract

AI Safety Fundamental Final Project

12 minute read

Investigating Local circuit in the feature basis

AI Control with Mechanistic Interpretability

less than 1 minute read

I’m excited to share our recent paper on AI Control through Mechanistic Interpretability approaches.

Towards a Probabilistic Disentanglement of Transformer Activations Part 2

6 minute read

This is the second post in our series dedicated to exploring methods around dictionary learning, and the possibility of a probabilistic take on the disentang...

Towards a Probabilistic Disentanglement of Transformer Activations Part 1

9 minute read

This is the first of a series of posts where I lay down some intuitions and best practices on how to perform probabilistic dictionary learning on Transformer...

Word Embeddings: A Comprehensive Guide Part 2

6 minute read

Word Embeddings: A Comprehensive Guide Part 2

Word Embeddings: A Comprehensive Guide Part 1

4 minute read

Table of Contents

Balanced Sentence Part 1

3 minute read

Balanced Sentence in GPT-2 Part 1

Introduction to mechanistic Interpretability

4 minute read

In the second post on my blog, I will present a brief introduction to the main goals and methods of Mechanistic Interpretability (MI), as well as its history.

My first post

1 minute read

Welcome to my blog