Research notes, plotted.

§ BLOG · 22ENTRIES · NEWEST FIRST
APR 24breakthrough7 min
Block-Diagonal Transformers Win 8 of 17 Downstream Tasks
APR 24breakthrough7 min
The Compositional Advantage: Why Structure Helps Reasoning
APR 24pivot7 min
The EML Paradox: Better Language Model, Worse Symbolic Regression
APR 24breakthrough6 min
What Frozen Probes Reveal About BD4 Representations
APR 24breakthrough8 min
JEPA Meets Group Theory: Perfect Accuracy on S_4 Composition
APR 24kill6 min
When Your Results Are Real But Your Sample Size Is Not
APR 24fundamentals8 min
Structured Transformers in 2026: Where FFN Research Stands
APR 20perspectives5 min
The Case for Negative Results in ML Research
APR 20tutorials10 min
Conformal Prediction for Neural Networks: A Practical Introduction
APR 20fundamentals8 min
What Happens Inside a Transformer FFN Layer? A Visual Guide
APR 20fundamentals12 min
Representation Theory for Machine Learning Engineers
APR 19fundamentals7 min
Block-Diagonal Matrices: Why Sparsity Structure Matters
APR 19tutorials8 min
Merkle Trees for Neural Network Verification
APR 18literature-review9 min
Multi-Hop Reasoning in Language Models: What Works and What Doesn't
APR 18research-notes6 min
Scaling Experiments with Minimal Infrastructure
APR 18fundamentals6 min
Why Feed-Forward Layers Matter More Than You Think
APR 17literature-review10 min
Equivariant Neural Networks: Beyond Data Augmentation
APR 17research-notes7 min
Statistical Rigor in Small-Scale ML Experiments
APR 15fundamentals7 min
Group Theory Meets Machine Learning: An Introduction
APR 10fundamentals6 min
The Grokking Phenomenon: When Neural Networks Suddenly Generalize
APR 05literature-review7 min
Structured Matrices in Neural Networks: A Survey
MAR 28perspectives7 min
Towards Verifiable AI: Formal Guarantees for Neural Network Outputs