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

Mechanistic interpretability (MI) is an emerging sub-field of interpretability that seeks to
understand a neural network model by reverse-engineering its internal computations …

How do language models use information provided as context when generating a
response? Can we infer whether a particular generated statement is actually grounded in …

Automated interpretability research has recently attracted attention as a potential research
direction that could scale explanations of neural network behavior to large models. Existing …

Information flows by routes inside the network via mechanisms implemented in the model.
These routes can be represented as graphs where nodes correspond to token …

How does the internal computation of a machine learning model transform inputs into
predictions? In this paper, we introduce a task called component modeling that aims to …

While artificial intelligence has the potential to process vast amounts of data, generate new
insights, and unlock greater productivity, its widespread adoption may entail unforeseen …

Recent work has found that sparse autoencoders (SAEs) are an effective technique for
unsupervised discovery of interpretable features in language models'(LMs) activations, by …

Most currently deployed large language models (LLMs) undergo continuous training or
additional finetuning. By contrast, most research into LLMs' internal mechanisms focuses on …

Multiple-choice question answering (MCQA) is a key competence of performant transformer
language models that is tested by mainstream benchmarks. However, recent evidence …