JoshEngels / SAE-Dark-MatterLinks
Code for our paper "Decomposing The Dark Matter of Sparse Autoencoders"
☆24Updated 8 months ago
Alternatives and similar repositories for SAE-Dark-Matter
Users that are interested in SAE-Dark-Matter are comparing it to the libraries listed below
Sorting:
- Delphi was the home of a temple to Phoebus Apollo, which famously had the inscription, 'Know Thyself.' This library lets language models …☆215Updated this week
- Stanford NLP Python library for benchmarking the utility of LLM interpretability methods☆136Updated 3 months ago
- Steering vectors for transformer language models in Pytorch / Huggingface☆125Updated 7 months ago
- Code to enable layer-level steering in LLMs using sparse auto encoders☆26Updated 2 weeks ago
- Code for reproducing our paper "Not All Language Model Features Are Linear"☆80Updated 10 months ago
- Code for my NeurIPS 2024 ATTRIB paper titled "Attribution Patching Outperforms Automated Circuit Discovery"☆41Updated last year
- Sparse Autoencoder Training Library☆54Updated 5 months ago
- Code for the ICLR 2024 paper "How to catch an AI liar: Lie detection in black-box LLMs by asking unrelated questions"☆72Updated last year
- ☆124Updated last year
- Improving Steering Vectors by Targeting Sparse Autoencoder Features☆24Updated 10 months ago
- ☆91Updated last year
- Code repo for the model organisms and convergent directions of EM papers.☆29Updated 2 weeks ago
- Investigating the generalization behavior of LM probes trained to predict truth labels: (1) from one annotator to another, and (2) from e…☆28Updated last year
- Create feature-centric and prompt-centric visualizations for sparse autoencoders (like those from Anthropic's published research).☆221Updated 9 months ago
- ☆32Updated 7 months ago
- ☆45Updated 8 months ago
- Steering Llama 2 with Contrastive Activation Addition☆187Updated last year
- Improving Alignment and Robustness with Circuit Breakers☆235Updated last year
- Code for reproducing our paper "Are Sparse Autoencoders Useful? A Case Study in Sparse Probing"☆29Updated 6 months ago
- Materials for "Multi-property Steering of Large Language Models with Dynamic Activation Composition"☆12Updated 10 months ago
- A library for efficient patching and automatic circuit discovery.☆77Updated 2 months ago
- ☆242Updated last year
- Open source replication of Anthropic's Crosscoders for Model Diffing☆59Updated 11 months ago
- PaCE: Parsimonious Concept Engineering for Large Language Models (NeurIPS 2024)☆40Updated 11 months ago
- ☆125Updated 2 weeks ago
- Code release for "Debating with More Persuasive LLMs Leads to More Truthful Answers"☆117Updated last year
- A benchmark for mechanistic discovery of circuits in Transformers☆16Updated 9 months ago
- Contains random samples referenced in the paper "Sleeper Agents: Training Robustly Deceptive LLMs that Persist Through Safety Training".☆118Updated last year
- ☆174Updated 10 months ago
- LLM experiments done during SERI MATS - focusing on activation steering / interpreting activation spaces☆97Updated 2 years ago