UOUO: Uncontextualized Uncommon Objects for Measuring Knowledge Horizons of Vision Language Models

TL;DR

This paper introduces the UOUO benchmark to evaluate vision-language models' ability to recognize rare and specialized objects, revealing that smaller models struggle with long-tail data despite performing well on common tasks.

Contribution

The paper presents a new benchmark for testing VLMs on uncommon objects and provides a scalable pipeline for data collection and cleaning.

Findings

Smaller VLMs underperform on rare object tasks.

Benchmark highlights limitations of current models on long-tail data.

Proposed pipeline ensures high-quality, challenging test instances.

Abstract

Smaller-scale Vision-Langauge Models (VLMs) often claim to perform on par with larger models in general-domain visual grounding and question-answering benchmarks while offering advantages in computational efficiency and storage. However, their ability to handle rare objects, which fall into the long tail of data distributions, is less understood. To rigorously evaluate this aspect, we introduce the "Uncontextualized Uncommon Objects" (UOUO) benchmark. This benchmark focuses on systematically testing VLMs with both large and small parameter counts on rare and specialized objects. Our comprehensive analysis reveals that while smaller VLMs maintain competitive performance on common datasets, they significantly underperform on tasks involving uncommon objects. We also propose an advanced, scalable pipeline for data collection and cleaning, ensuring the UOUO benchmark provides high-quality, challenging instances. These findings highlight the need to consider long-tail distributions when assessing the true capabilities of VLMs.