Tutorials

Walkthrough

Learn the basics

Activation Patching

Causal intervention

Attribution Patching

Approximate patching

Logit Lens

Decode activations

Future Lens

Probe future tokens

Function Vectors

Lambdas

Dictionary Learning

Sparse autoencoders

• Attribution Patching
• Setup (Ignore)
• 1️⃣ IOI Patching Setup
• 2️⃣ Attribution Patching Over Components
• 3️⃣ Attribution Patching Over Position
• Function Vectors
• Future Lens
• Activation Patching
• Logit Lens 🔎
• Setup (Ignore)
• GPT2 Model Architecture
• Dictionary Learning
• Walkthrough
• 1 First, let’s start small
• 2 Bigger
• 3 I thought you said huge models?
• Getting Involved!