Sae Python Packages

glassbox-mech-interp

Open-source EU AI Act Annex IV documentation toolkit. Mechanistic interpretability + circuit discovery for transformers. One function call generates a structured, hash-chained evidence package.

interp-lab

An open LLM interpretability tool that helps researchers and AI agents identify LLM activations and features and make Sparse Auto Encoders as needed.

glassbox-mcp

Open-source EU AI Act Annex IV documentation toolkit. Mechanistic interpretability + circuit discovery for transformers. One function call generates a structured, hash-chained evidence package.

openinterp

openinterp — Python SDK + CLI. FabricationGuard hallucination probe + ProbeBench leaderboard + Atlas search + Trace generation. pip install openinterp

prefscope

An open-source framework for analyzing behavioral patterns in post-training datasets and language-model responses using sparse autoencoders and model-based interpretation.

gated-sae-tf

🧩 Sparse Gated Autoencoders for TensorFlow & Keras — dictionary learning & mechanistic interpretability, batteries included. The gated SAE with the full training recipe + sparsity/sharpness reports and feature galleries. pip install gated-sae-tf

mechreward

Mechanistic interpretability as reward signal for RL training of LLMs — SAE features + GRPO + anti-Goodhart framework