Interpretability & Mechanistic Analysis

Interpretability & Mechanistic Analysis

dynaxx on the hidden graph: interpretability via circuit-level abstraction

If you have ever stared at a model's attention patterns and felt no closer to understanding why it made a specific prediction, you are not alone. Raw ...

Interpretability & Mechanistic Analysis

The Dynaxx Mechanistic Audit: Tracing Causal Paths in Production Models

This comprehensive guide explores the Dynaxx Mechanistic Audit, a rigorous methodology for tracing causal paths in production models. Unlike conventio...

Interpretability & Mechanistic Analysis

The dynaxx map: tracing feature superposition to latent circuit topology

Introduction: The Problem of Feature SuperpositionWhen we inspect the internal representations of a deep neural network, we often find that individual...

Interpretability & Mechanistic Analysis

Dynaxx on the Circuitry Frontier: Mechanistic Probes for Superposition and Polysemanticity

Introduction: The Black Box is a MirageIn my decade of working with deep learning systems, first in academic research and now leading the Dynaxx resea...

Interpretability & Mechanistic Analysis

The Dynaxx Dissection: Reverse-Engineering In-Context Learning as Implicit Gradient Descent

In-context learning (ICL) lets large language models adapt to new tasks with just a few prompt examples, no weight updates. For interpretability resea...