An adversarial example library for constructing attacks, building defenses, and benchmarking both

Book about interpretable machine learning

The hub for EleutherAI's work on interpretability and learning dynamics

Discovering Interpretable GAN Controls [NeurIPS 2020]

Takagi and Nishimoto, CVPR 2023

TruthfulQA: Measuring How Models Imitate Human Falsehoods

Emergent world representations: Exploring a sequence model trained on a synthetic task

Training pipeline for end-to-end self-driving with Comma AI's Openpilot. WIP

tomsup 👍 Theory of Mind Simulation using Python. A package that allows for easy agent-based modelling of recursive Theory of Mind

WMDP is a LLM proxy benchmark for hazardous knowledge in bio, cyber, and chemical security. We also release code for RMU, an unlearning method which reduces LLM performance on WMDP while retaining …

Machine Learning for Alignment Bootcamp

Tools for exploring Transformer neuron behaviour, including input pruning and diversification.

✱ Understanding the underlying learning dynamics of simple tasks in Transformer networks

Mechanistic Interpretability Tutorials, Results and research log as I learn from publicly available research, and experimentation.

🔥 A repository for collecting cyberdefense thoughts, books, and documents about AI cyberdefense

Code templates to get started as an AI psychologist

This repository contains code for the Democracy x AI Hackathon by Apart Research

A tool for exploring EA Forum and LessWrong

Alignment Jam - Interpretability Hackaton

An easy straightforward tutorial for finetuning a Danish BERT using simpletransformers

🧠 EEG classification in a P300 typing task from the BR41N.IO hackathon 2021

Interpretability Hackathon 2.0 entry

Tools for exploring Transformer neuron behaviour, including input pruning and diversification.