Taking Shortcuts for Categorical VQA Using Super Neurons

TL;DR

This work finds 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 and robustly improves the classification performance while achieving a speedup of up to 5.10x.

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.

Figures (8)

• Figure 1: Overview of our approach. Our training-free scheme uncovers Super Neurons (SNs) via probing data. They robustly outperform the base model on a variety of categorical VQA datasets. As a byproduct, they enable extreme early exiting at the first layer of the LLM on the first generated token. Colored boxes in the LLM represent SNs for their data types.
• Figure 2: Comparison between SAVs and ours. We show architectural divergences in \ref{['fig:ours_vs_savs']} and resulting search spaces of the two approaches for LLaVA-v1.5-7b in \ref{['tab:search_space']}.
• Figure 3: Empirical analysis of SNs.\ref{['fig:act_thresh_range']} shows the maximum accuracy on the probing set with respect to different $\alpha$. We evaluate $\alpha$ over the range $\alpha \in [-3, 3]$. The maximum accuracy peaks around $\alpha = 0$. \ref{['fig:neurons_per_layer']} records the number of found SNs that obtain a better accuracy than the model in each layer. We use LLaVA-v1.5-7b on VizWiz for both figures.
• Figure 4: Agreement rate with respect to different SN$t$. We compute AR on Pope using LLaVA-v1.5-7b. At lower accuracy, SNs largely agree with the model prediction. However, for SNs to obtain better results than the model, they have to disagree on some answers.
• Figure 5: Super Neurons performances with respect to different probing set sizes. We compute accuracy, precision, recall and F1 on diverse benchmarks using LLaVA-v1.5-7b. Dashed lines (colors matching their respective datasets) indicate the performance of the vanilla model. Overall, more data leads to performance improvements.