noanabeshima / matryoshka-saes
☆16Updated 4 months ago
Alternatives and similar repositories for matryoshka-saes:
Users that are interested in matryoshka-saes are comparing it to the libraries listed below
- Sparse Autoencoder Training Library☆48Updated 5 months ago
- ☆31Updated 3 months ago
- Improving Steering Vectors by Targeting Sparse Autoencoder Features☆17Updated 5 months ago
- ☆22Updated 2 months ago
- ☆42Updated last year
- Providing the answer to "How to do patching on all available SAEs on GPT-2?". It is an official repository of the implementation of the p…☆11Updated 3 months ago
- A library for efficient patching and automatic circuit discovery.☆63Updated this week
- Redwood Research's transformer interpretability tools☆14Updated 3 years ago
- ☆26Updated last year
- Universal Neurons in GPT2 Language Models☆27Updated 10 months ago
- ☆36Updated last month
- Evaluate interpretability methods on localizing and disentangling concepts in LLMs.☆43Updated 6 months ago
- ☆13Updated last year
- A small package implementing some useful wrapping around nnsight☆12Updated last week
- ☆36Updated 5 months ago
- ☆19Updated last year
- Attribution-based Parameter Decomposition☆18Updated this week
- Simple and scalable tools for data-driven pretraining data selection.☆22Updated 2 months ago
- Code for reproducing our paper "Not All Language Model Features Are Linear"☆73Updated 4 months ago
- Open source replication of Anthropic's Crosscoders for Model Diffing☆52Updated 5 months ago
- Official repository for our paper, Transformers Learn Higher-Order Optimization Methods for In-Context Learning: A Study with Linear Mode…☆16Updated 5 months ago
- NeuroSurgeon is a package that enables researchers to uncover and manipulate subnetworks within models in Huggingface Transformers☆41Updated 2 months ago
- ☆85Updated 2 weeks ago
- Code Release for "Broken Neural Scaling Laws" (BNSL) paper☆58Updated last year
- Code and Data Repo for the CoNLL Paper -- Future Lens: Anticipating Subsequent Tokens from a Single Hidden State☆18Updated last year
- Code for my NeurIPS 2024 ATTRIB paper titled "Attribution Patching Outperforms Automated Circuit Discovery"☆30Updated 10 months ago
- Investigating the generalization behavior of LM probes trained to predict truth labels: (1) from one annotator to another, and (2) from e…☆26Updated 11 months ago