Understanding AI systems' inner workings is critical for ensuring value alignment and safety.
This review explores mechanistic interpretability: reverse engineering the computational …

Sparse autoencoders (SAEs) are an unsupervised method for learning a sparse
decomposition of a neural network's latent representations into seemingly interpretable …

The remarkable performance of large language models (LLMs) in content generation,
coding, and common-sense reasoning has spurred widespread integration into many facets …

The rise of the term" mechanistic interpretability" has accompanied increasing interest in
understanding neural models--particularly language models. However, this jargon has also …

Mechanistic interpretability aims to understand the computational mechanisms underlying
neural networks' capabilities in order to accomplish concrete scientific and engineering …

The greatest ambition of mechanistic interpretability is to completely rewrite deep neural
networks in a format that is more amenable to human understanding, while preserving their …

Multi-Agent Deep Reinforcement Learning (MADRL) was proven efficient in solving complex
problems in robotics or games, yet most of the trained models are hard to interpret. While …

A common goal of mechanistic interpretability is to decompose the activations of neural
networks into features: interpretable properties of the input computed by the model. Sparse …

While induction is considered a key mechanism for in-context learning in LLMs,
understanding its precise circuit decomposition beyond toy models remains elusive. Here …

We introduce a new approach to systematically map features discovered by sparse
autoencoder across consecutive layers of large language models, extending earlier work …