bartbussmann / BatchTopK
Implementation of the BatchTopK activation function for training sparse autoencoders (SAEs)
☆32Updated 6 months ago
Alternatives and similar repositories for BatchTopK:
Users that are interested in BatchTopK are comparing it to the libraries listed below
- ☆78Updated last week
- ☆33Updated last month
- ☆26Updated 2 months ago
- Open source replication of Anthropic's Crosscoders for Model Diffing☆46Updated 5 months ago
- ☆61Updated 4 months ago
- Delphi was the home of a temple to Phoebus Apollo, which famously had the inscription, 'Know Thyself.' This library lets language models …☆163Updated this week
- ☆17Updated 11 months ago
- Code for reproducing our paper "Not All Language Model Features Are Linear"☆72Updated 4 months ago
- ☆121Updated last year
- Sparse Autoencoder Training Library☆43Updated 4 months ago
- Multi-Layer Sparse Autoencoders (ICLR 2025)☆21Updated last month
- ☆82Updated 7 months ago
- This repository contains the code used for the experiments in the paper "Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity…☆25Updated last year
- A library for efficient patching and automatic circuit discovery.☆59Updated last month
- Using sparse coding to find distributed representations used by neural networks.☆224Updated last year
- ☆9Updated 3 months ago
- ☆23Updated last month
- ☆32Updated 4 months ago
- Code for NeurIPS 2024 Spotlight: "Scaling Laws and Compute-Optimal Training Beyond Fixed Training Durations"☆71Updated 4 months ago
- ☆13Updated 4 months ago
- Steering Llama 2 with Contrastive Activation Addition☆131Updated 10 months ago
- Sparse and discrete interpretability tool for neural networks☆59Updated last year
- ☆90Updated last month
- Create feature-centric and prompt-centric visualizations for sparse autoencoders (like those from Anthropic's published research).☆190Updated 3 months ago
- Code for our paper "Decomposing The Dark Matter of Sparse Autoencoders"☆22Updated last month
- Simple and scalable tools for data-driven pretraining data selection.☆17Updated last month
- Code for my NeurIPS 2024 ATTRIB paper titled "Attribution Patching Outperforms Automated Circuit Discovery"☆30Updated 9 months ago
- ☆21Updated 6 months ago
- Official implementation of "BERTs are Generative In-Context Learners"☆26Updated 2 weeks ago
- ☆150Updated 3 weeks ago