evan-lloyd / graphpatchLinks
graphpatch is a library for activation patching on PyTorch neural network models.
☆20Updated 7 months ago
Alternatives and similar repositories for graphpatch
Users that are interested in graphpatch are comparing it to the libraries listed below
Sorting:
- MishformerLens intends to be a drop-in replacement for TransformerLens that AST patches HuggingFace Transformers rather than implementing…☆10Updated 11 months ago
- ☆127Updated last year
- Create feature-centric and prompt-centric visualizations for sparse autoencoders (like those from Anthropic's published research).☆220Updated 9 months ago
- Sparse Autoencoder Training Library☆54Updated 4 months ago
- Erasing concepts from neural representations with provable guarantees☆234Updated 8 months ago
- ☆242Updated 11 months ago
- Mechanistic Interpretability Visualizations using React☆289Updated 9 months ago
- ☆81Updated 7 months ago
- Notebooks accompanying Anthropic's "Toy Models of Superposition" paper☆129Updated 3 years ago
- Steering vectors for transformer language models in Pytorch / Huggingface☆124Updated 7 months ago
- 🧠 Starter templates for doing interpretability research☆74Updated 2 years ago
- Open source replication of Anthropic's Crosscoders for Model Diffing☆59Updated 11 months ago
- ☆276Updated last year
- Extract full next-token probabilities via language model APIs☆248Updated last year
- ☆142Updated 2 weeks ago
- Mechanistic Interpretability for Transformer Models☆51Updated 3 years ago
- ☆34Updated last year
- Utilities for the HuggingFace transformers library☆72Updated 2 years ago
- ☆71Updated 3 weeks ago
- ☆345Updated last month
- Delphi was the home of a temple to Phoebus Apollo, which famously had the inscription, 'Know Thyself.' This library lets language models …☆211Updated last week
- ☆109Updated 7 months ago
- How do transformer LMs encode relations?☆53Updated last year
- ☆51Updated 2 months ago
- PyTorch and NNsight implementation of AtP* (Kramar et al 2024, DeepMind)☆19Updated 8 months ago
- Sparse Autoencoder for Mechanistic Interpretability☆267Updated last year
- A collection of different ways to implement accessing and modifying internal model activations for LLMs☆19Updated 11 months ago
- Code for my NeurIPS 2024 ATTRIB paper titled "Attribution Patching Outperforms Automated Circuit Discovery"☆41Updated last year
- Understand and test language model architectures on synthetic tasks.☆226Updated this week
- Keeping language models honest by directly eliciting knowledge encoded in their activations.☆209Updated this week