Awesome LLM Interpretability
A curated list of tools, papers, articles and communities for understanding how large language models work internally.
Awesome LLM Interpretability curates tools, papers, articles and communities focused on understanding the internal workings of large language models. Its tools section lists projects such as the Learning Interpretability Tool for visualizing classification, regression and generative models, Comgra and Pythia for analyzing neural networks during training, Phoenix for evaluating and troubleshooting LLM, computer-vision and NLP models, sparse-autoencoder and attention-analysis tooling, and libraries such as TransformerLens for mechanistic interpretability of generative language models.
Beyond tools, the list organizes papers, articles and interpretability-focused research groups into a single table of contents, giving a starting point for someone new to the field to find the most relevant prior work. It is maintained as a plain, community-editable list on GitHub.