In plain words: A vision-language model gets tools to make and edit images, find the pictures that most excite a neuron, and summarize results, letting it run its own experiments to explain how other vision models work. Its neuron descriptions matched those of expert human researchers.
Abstract
This paper describes MAIA, a Multimodal Automated Interpretability Agent. MAIA is a system that uses neural models to automate neural model understanding tasks like feature interpretation and failure mode discovery. It equips a pre-trained vision-language model with a set of tools that support iterative experimentation on subcomponents of other models to explain their behavior. These include tools commonly used by human interpretability researchers: for synthesizing and editing inputs, computing maximally activating exemplars from real-world datasets, and summarizing and describing experimental results. Interpretability experiments proposed by MAIA compose these tools to describe and explain system behavior. We evaluate applications of MAIA to computer vision models. We first characterize MAIA's ability to describe (neuron-level) features in learned representations of images. Across several trained models and a novel dataset of synthetic vision neurons with paired ground-truth descriptions, MAIA produces descriptions comparable to those generated by expert human experimenters. We then show that MAIA can aid in two additional interpretability tasks: reducing sensitivity to spurious features, and automatically identifying inputs likely to be mis-classified.
Tamar Rott Shaham, Sarah Schwettmann, Franklin Wang, Achyuta Rajaram, Evan Hernandez, Jacob Andreas, Antonio Torralba
arXiv:2404.14394 · cs.AI, cs.CL, cs.CV · submitted Apr 22, 2024 · updated Feb 11, 2025
abstract · pdf · html · 25 pages, 13 figures
For folks who are more familiar with this branch of literature, given the above, why is this a fruitful line of inquiry? Isn't this akin to stacking turtles on top of each other?