A library for making RepE control vectors

A complete end-to-end pipeline for LLM interpretability with sparse autoencoders (SAEs) using Llama 3.2, written in pure PyTorch and fully reproducible.

This repository collects all relevant resources about interpretability in LLMs

Implementation of the stacked denoising autoencoder in Tensorflow

Pivotal Token Search

Official Code for Paper: Beyond Matryoshka: Revisiting Sparse Coding for Adaptive Representation

Pytorch implementations of various types of autoencoders

Evaluate interpretability methods on localizing and disentangling concepts in LLMs.

SANSA - sparse EASE for millions of items

Sparse Embedding Compression for Scalable Retrieval in Recommender Systems

Explore visualization tools for understanding Transformer-based large language models (LLMs)

Multi-Layer Sparse Autoencoders (ICLR 2025)

Tensorflow Examples

Official Triton kernels for TopK and HierarchicalTopK Sparse Autoencoder decoders.

Sparse Auto Encoder and regular MNIST classification with mini batch's

Experiments with Adversarial Autoencoders using Keras

Real-time 3D visualisation of SAE feature activations inside GPT-2, token by token

Sparse Autoencoders (SAE) vs CLIP fine-tuning fun.

A tiny easily hackable implementation of a feature dashboard.

[NeurIPS 2025] This is the official repository for VL-SAE: Interpreting and Enhancing Vision-Language Alignment with a Unified Concept Set