ckkissane / sae-transfer
Code to reproduce key results accompanying "SAEs (usually) Transfer Between Base and Chat Models"
☆11Updated 9 months ago
Alternatives and similar repositories for sae-transfer:
Users that are interested in sae-transfer are comparing it to the libraries listed below
- Sparse Autoencoder Training Library☆48Updated 5 months ago
- Universal Neurons in GPT2 Language Models☆27Updated 10 months ago
- ☆36Updated 2 months ago
- A library for efficient patching and automatic circuit discovery.☆63Updated this week
- 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
- Code to enable layer-level steering in LLMs using sparse auto encoders☆16Updated 7 months ago
- Algebraic value editing in pretrained language models☆63Updated last year
- ☆82Updated 8 months ago
- ☆19Updated last year
- Code for reproducing our paper "Not All Language Model Features Are Linear"☆73Updated 4 months ago
- Code for my NeurIPS 2024 ATTRIB paper titled "Attribution Patching Outperforms Automated Circuit Discovery"☆31Updated 10 months ago
- ☆13Updated last year
- PyTorch and NNsight implementation of AtP* (Kramar et al 2024, DeepMind)☆18Updated 3 months ago
- ☆104Updated 5 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
- Stanford NLP Python library for benchmarking the utility of LLM interpretability methods☆70Updated 3 weeks ago
- Open source replication of Anthropic's Crosscoders for Model Diffing☆52Updated 5 months ago
- Evaluate interpretability methods on localizing and disentangling concepts in LLMs.☆43Updated 6 months ago
- ☆9Updated this week
- ☆34Updated last year
- Multi-Layer Sparse Autoencoders (ICLR 2025)☆22Updated 2 months ago
- A small package implementing some useful wrapping around nnsight☆12Updated last week
- Code for the ICLR 2024 paper "How to catch an AI liar: Lie detection in black-box LLMs by asking unrelated questions"☆69Updated 10 months ago
- ☆26Updated last year
- Code and Data Repo for the CoNLL Paper -- Future Lens: Anticipating Subsequent Tokens from a Single Hidden State☆18Updated last year
- ☆22Updated 2 months ago
- ☆29Updated last week
- Attribution-based Parameter Decomposition☆18Updated this week
- Steering vectors for transformer language models in Pytorch / Huggingface☆95Updated 2 months ago
- 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