Total: 1
Multimodal Large Language Models (MLLMs) have recently achieved promising performance on visual question answering (VQA)---a fundamental task affecting various downstream applications and domains. Given MLLMs' potential integration into many critical VQA applications, it is important to understand the limits of their perception. In this work, we study whether MLLMs can perceive small details as well as large details in images. In particular, we observe that their accuracy in answering visual questions is very sensitive to the size of the visual subject of the question. We further show that this effect is causal by observing that human visual cropping can significantly mitigate this sensitivity. Next, we study the attention patterns of MLLMs when answering visual questions, and intriguingly find that they consistently know where to look, even when they provide the wrong answer. Based on these findings, we then construct automatic visual cropping methods that leverage the internal knowledge of any MLLM itself, in the form of attention and gradient maps, to help it better perceive the small visual subject of any question. We study our proposed methods on two MLLMs and seven visual question answering benchmarks, and show that they can significantly improve MLLMs accuracy without requiring any training. Our findings suggest that MLLMs should be used with caution in detail-sensitive applications, and that visual cropping is a promising direction to improve their performance.