Taking Shortcuts for Categorical VQA Using Super Neurons

TL;DR

This paper introduces Super Neurons, a method that uses scalar activations from early layers of vision-language models to create accurate classifiers, enabling faster inference and improved performance on visual tasks.

Contribution

It proposes using scalar activations (Super Neurons) from early model layers for classification, offering a training-free, speed-boosting alternative to attention-based methods.

Findings

Super Neurons improve classification accuracy.

Enables early exiting from the first model layer.

Achieves up to 5.10x speedup.

Abstract

Sparse Attention Vectors (SAVs) have emerged as an excellent training-free alternative to supervised finetuning or low-rank adaptation to improve the performance of Vision Language Models (VLMs). At their heart, SAVs select a few accurate attention heads for a task of interest and use them as classifiers, rather than relying on the model's prediction. In a similar spirit, we find that directly probing the raw activations of the VLM, in the form of scalar values, is sufficient to yield accurate classifiers on diverse visually grounded downstream tasks. Shifting focus from attention vectors to scalar activations dramatically increases the search space for accurate parameters, allowing us to find more discriminative neurons immediately from the first generated token. We call such activations Super Neurons (SNs). In this probing setting, we discover that enough SNs appear in the shallower layers of the large language model to allow for extreme early exiting from the first layer of the model at the first generated token. Compared to the original network, SNs robustly improve the classification performance while achieving a speedup of up to 5.10x.