jbloomAus

SAELens

Training Sparse Autoencoders on Language Models

DecisionTransformerInterpretability

Interpreting how transformers simulate agents performing RL tasks

alphabetical_probe

Experimental code which trains 26 linear probes to detect the presence of alphabetic letters in GPT-J token strings, given their embeddings. Exploring the resulting vector arithmetic and its impact on GPT-J spelling abilities

SAEDashboard

toy_model_interpretability

I'd like to start playing around with toy models to better understand results in recent papers.

TransformerLens

arena-v1

arena-v1-ldn

ARENA_2.0

I'm teaching ARENA 2.0 and providing students with direction on careers and personal development.

ARENA_2.0-RLHF

Preparing content for the ARENA RLHF day.

SparseAutoencoderSuperposition

babyai

BabyAI platform. A testbed for training agents to understand and execute language commands.

Backwards

Exploring-2L-SAE

geom_median

Fast and differentiable geometric median, a multivariate median analogue. Install with `pip install geom-median`

Minigrid

Simple and easily configurable grid world environments for reinforcement learning

Module-1

Module 1 - Autodifferentiation

post--memory-dt-features

protein-inference

A python package for protein inference in Mass Spectrometric data analysis.

rust_cli_project

I'm teaching myself Rust.

rust_text_editor

Learning by doing with Rust. Following along the Hecto tutorial https://www.philippflenker.com/hecto/

sparse_autoencoder

Sparse Autoencoder for Mechanistic Interpretability