Marcin Piotrowski
software engineer
☰ Menu

LLM-Scope — Visualizing What Happens Inside LLM

LLM mechanistic interpretability PyTorch transformers attention visualization Python

A tool that shows token by token what happens inside Qwen3-0.6B while it reads a prompt and generates a response: MLP neuron activations, attention, a logit lens, and a full 'token journey' through every transformer layer.

Read more →
LLM-Scope — Visualizing What Happens Inside LLM