-
Why Some Puzzles Are Hard for LLMs: A Framework for Cognitive Load
-
The Information Efficiency of Linguistic Structure
-
Swimming with the Whale: A Model for Market Maker Price Manipulation
-
Tips on Meta-Prompting
-
The Energetic Cost of Computation: A Stochastic Thermodynamics Primer
-
Hallucination and Context Filling: A Toy Model
-
Mechanistic Interpretability: Part 8 - Circuit Tracing and Insights from SOTA Models
-
Mechanistic Interpretability: Part 7 - Induction Heads: The Mechanics of In-Context Learning
-
Mechanistic Interpretability: Part 6 - Neural Network Circuits: Taxonomy and Attention Patterns
-
Mechanistic Interpretability: Part 5 - Validating Learned Features and Circuits
-
Mechanistic Interpretability: Part 4 - Mathematical Framework for Transformer Circuits
-
Mechanistic Interpretability: Part 3 - The Spectrum of Polysemanticity and Monosemanticity
-
Mechanistic Interpretability: Part 2 - The Superposition Hypothesis and Dictionary Learning
-
Mechanistic Interpretability: Part 1 - Foundations and the Circuits Paradigm
-
Scaling Law of LLM Hallucination
-
Preserving Curvature while Smoothing via Anisotropic Diffusion
-
Stable Distribution with Minimal Information
About Me
I'm interested in physics, maths, finance, and anything in-between. This blog contains short notes on these topics, heavily co-authored by LLM.