Transformer Circuits 6: Where Does This Leave Us?
Reading the Framework paper, part 6. What the framework secured and what it did not: the wall called MLP, and the follow-up research this paper announces.
Posts
Transformer Circuits 5: Two-Layer Models and Induction Heads
Reading the Framework paper, part 5. With two layers, heads compose with heads: Q-, K-, and V-composition, and the paper’s biggest discovery, the mechanism of the induction head.
Transformer Circuits 4: One-Layer Models and Skip-Trigrams
Reading the Framework paper, part 4. Fully expanding a one-attention-layer model into paths: skip-trigrams, copying heads and positive eigenvalues, and the characteristic bugs the factored structure produces.
Transformer Circuits 3: Zero-Layer Transformers
Reading the Framework paper, part 3. What does a transformer with no attention at all learn? Bigram statistics, and that answer turns out to be the key to the direct path in large models.
Transformer Circuits 2: The Transformer, Redrawn
Reading the Framework paper, part 2. Redrawing the residual stream as a communication channel and attention heads as independent information movers, through virtual weights and the QK/OV split.
Transformer Circuits 1: Summary of Results
Reading the Framework paper, part 1. Treating tiny attention-only models as model organisms for fully reverse engineering transformers, and a map of the key results.
Mechanistic Interpretability: Free Materials to Start
Reverse-engineering neural networks, circuit by circuit. Anthropic’s Transformer Circuits thread and a few companion resources are all free. Where to start, to open this series.
Steering GPT-2's Emotions with Sparse Autoencoders
Finding emotion-related features in GPT-2 using OpenAI’s pretrained SAE, then training one from scratch. Feature patching turns ‘good person’ into ‘shit’.
MacKay's Information Theory: The Free Book and the Lectures
David MacKay’s classic ties information theory, inference, and machine learning into one story, and both the book and the lecture videos are free. Where to get them, to open this series.
K-means, GMM, EM: Three Nested Russian Dolls of Clustering
K-means is actually an extreme case of GMM, and GMM is the canonical application of the EM algorithm. How these three connect within a single framework, and how information geometry explains the relationship.