Vision-Language Models under Cultural and Inclusive Considerations

TL;DR

This paper introduces a culture-centric evaluation benchmark for vision-language models, highlighting their performance and challenges in diverse cultural contexts, especially for assisting visually impaired users.

Contribution

It creates a new culturally diverse evaluation dataset and benchmark, and assesses the reliability of VLMs in inclusive, real-world scenarios.

Findings

State-of-the-art models show promising results

Challenges include hallucination and metric misalignment

Provides publicly available survey, data, and code

Abstract

Large vision-language models (VLMs) can assist visually impaired people by describing images from their daily lives. Current evaluation datasets may not reflect diverse cultural user backgrounds or the situational context of this use case. To address this problem, we create a survey to determine caption preferences and propose a culture-centric evaluation benchmark by filtering VizWiz, an existing dataset with images taken by people who are blind. We then evaluate several VLMs, investigating their reliability as visual assistants in a culturally diverse setting. While our results for state-of-the-art models are promising, we identify challenges such as hallucination and misalignment of automatic evaluation metrics with human judgment. We make our survey, data, code, and model outputs publicly available.