Response Wide Shut? Surprising Observations in Basic Vision Language Model Capabilities

TL;DR

This paper investigates the fundamental limitations of state-of-the-art vision-language models by designing comprehensive tests that reveal their weaknesses in basic visual understanding, comparing their responses to intermediate visual features.

Contribution

It introduces new probing methods that go beyond standard benchmarks to analyze the internal components and processing of VLMs, uncovering their specific shortcomings.

Findings

VLMs lack robustness in basic visual tasks

Intermediate visual features reveal hidden model limitations

Probing shows discrepancies between model responses and visual features

Abstract

Vision-language Models (VLMs) have emerged as general-purpose tools for addressing a variety of complex computer vision problems. Such models have been shown to be highly capable, but, at the same time, lacking some basic visual understanding skills. In this paper, we set out to understand the limitations of SoTA VLMs on fundamental visual tasks by constructing a series of tests that probe which components of design, specifically, may be lacking. Importantly, we go significantly beyond the current benchmarks, which simply measure the final performance of VLM response, by also comparing and contrasting it to the performance of probes trained directly on features obtained from the visual encoder, intermediate vision-language projection and LLM-decoder output. In doing so, we uncover shortcomings in VLMs and make a number of important observations about their capabilities, robustness and how they process visual information. We hope our insights will guide progress in further improving VLMs.