safety-research / safety-tooling
Inference API for many LLMs and other useful tools for empirical research
☆33Updated this week
Alternatives and similar repositories for safety-tooling:
Users that are interested in safety-tooling are comparing it to the libraries listed below
- ☆13Updated 2 weeks ago
- ControlArena is a suite of realistic settings, mimicking complex deployment environments, for running control evaluations. This is an alp…☆50Updated this week
- METR Task Standard☆146Updated 2 months ago
- Machine Learning for Alignment Bootcamp☆72Updated 2 years ago
- ☆10Updated 9 months ago
- Resources for skilling up in AI alignment research engineering. Covers basics of deep learning, mechanistic interpretability, and RL.☆209Updated last year
- ☆218Updated 6 months ago
- Mechanistic Interpretability Visualizations using React☆241Updated 4 months ago
- ☆31Updated 11 months ago
- ☆89Updated last month
- (Model-written) LLM evals library☆18Updated 4 months ago
- Tools for studying developmental interpretability in neural networks.☆88Updated 3 months ago
- Vivaria is METR's tool for running evaluations and conducting agent elicitation research.☆89Updated last week
- ☆36Updated 5 months ago
- Open source replication of Anthropic's Crosscoders for Model Diffing☆52Updated 5 months ago
- Steering vectors for transformer language models in Pytorch / Huggingface☆95Updated 2 months ago
- ☆161Updated 2 weeks ago
- Improving Steering Vectors by Targeting Sparse Autoencoder Features☆17Updated 5 months ago
- ☆121Updated last year
- Collection of evals for Inspect AI☆115Updated this week
- Delphi was the home of a temple to Phoebus Apollo, which famously had the inscription, 'Know Thyself.' This library lets language models …☆169Updated this week
- A toolkit for describing model features and intervening on those features to steer behavior.☆178Updated 5 months ago
- ☆12Updated 2 weeks ago
- ☆85Updated 2 weeks ago
- Create feature-centric and prompt-centric visualizations for sparse autoencoders (like those from Anthropic's published research).☆197Updated 4 months ago
- The nnsight package enables interpreting and manipulating the internals of deep learned models.☆548Updated this week
- ☆49Updated 3 months ago
- A library for efficient patching and automatic circuit discovery.☆63Updated this week
- Applying SAEs for fine-grained control☆17Updated 4 months ago
- LLM experiments done during SERI MATS - focusing on activation steering / interpreting activation spaces☆91Updated last year