StartLuxLabs/Massive-Activations-HLA

Experiments on whether language models can learn to detect read-only activation probes

Official PyTorch Implementation for Learning a Generative Meta-Model of LLM Activations, ICML 2026

Companion code for the global workspace interpretability paper

A library for mechanistic interpretability of GPT-style language models

Mapping out the "memory" of neural nets with data attribution

Explorations into the recently proposed Taylor Series Linear Attention

Official code for HiLS-Attention

ICA Lens: compact ICA-based interpretability tools for exploring LLM activations. Code release for the paper.