NLP models today strive for supporting multiple languages and modalities, improving accessibility for diverse users. In this paper, we evaluate their multilingual, multimodal capabilities by testing on a visual reasoning task. We observe that proprietary systems like GPT-4V obtain the best performance on this task now, but open models lag in comparison. Surprisingly, GPT-4V exhibits similar performance between English and other languages, indicating the potential for equitable system development across languages. Our analysis on model failures reveals three key aspects that make this task challenging: multilinguality, complex reasoning, and multimodality. To address these challenges, we propose three targeted interventions including a translate-test approach to tackle multilinguality, a visual programming approach to break down complex reasoning, and a novel method that leverages image captioning to address multimodality. Our interventions achieve the best open performance on this task in a zero-shot setting, boosting open model LLaVA by 13.4%, while also minorly improving GPT-4V's performance.

Pretrained large Vision-Language models have drawn considerable interest in recent years due to their remarkable performance. Despite considerable efforts to assess these models from diverse perspectives, the extent of visual cultural awareness in the state-of-the-art GPT-4V model remains unexplored. To tackle this gap, we extensively probed GPT-4V using the MaRVL benchmark dataset, aiming to investigate its capabilities and limitations in visual understanding with a focus on cultural aspects. Specifically, we introduced three visual related tasks, i.e. caption classification, pairwise captioning, and culture tag selection, to systematically delve into fine-grained visual cultural evaluation. Experimental results indicate that GPT-4V excels at identifying cultural concepts but still exhibits weaker performance in low-resource languages, such as Tamil and Swahili. Notably, through human evaluation, GPT-4V proves to be more culturally relevant in image captioning tasks than the original MaRVL human annotations, suggesting a promising solution for future visual cultural benchmark construction.

Most multilingual vision-and-language (V&L) research aims to accomplish multilingual and multimodal capabilities within one model. However, the scarcity of multilingual captions for images has hindered the development. To overcome this obstacle, we propose ICU, Image Caption Understanding, which divides a V&L task into two stages: a V&L model performs image captioning in English, and a multilingual language model (mLM), in turn, takes the caption as the alt text and performs cross-lingual language understanding. The burden of multilingual processing is lifted off V&L model and placed on mLM. Since the multilingual text data is relatively of higher abundance and quality, ICU can facilitate the conquering of language barriers for V&L models. In experiments on two tasks across 9 languages in the IGLUE benchmark, we show that ICU can achieve new state-of-the-art results for five languages, and comparable results for the rest.

The design of widespread vision-and-language datasets and pre-trained encoders directly adopts, or draws inspiration from, the concepts and images of ImageNet. While one can hardly overestimate how much this benchmark contributed to progress in computer vision, it is mostly derived from lexical databases and image queries in English, resulting in source material with a North American or Western European bias. Therefore, we devise a new protocol to construct an ImageNet-style hierarchy representative of more languages and cultures. In particular, we let the selection of both concepts and images be entirely driven by native speakers, rather than scraping them automatically. Specifically, we focus on a typologically diverse set of languages, namely, Indonesian, Mandarin Chinese, Swahili, Tamil, and Turkish. On top of the concepts and images obtained through this new protocol, we create a multilingual dataset for {M}ulticultur{a}l {R}easoning over {V}ision and {L}anguage (MaRVL) by eliciting statements from native speaker annotators about pairs of images. The task consists of discriminating whether each grounded statement is true or false. We establish a series of baselines using state-of-the-art models and find that their cross-lingual transfer performance lags dramatically behind supervised performance in English. These results invite us to reassess the robustness and accuracy of current state-of-the-art models beyond a narrow domain, but also open up new exciting challenges for the development of truly multilingual and multicultural systems.