Evaluating Linguistic Capabilities of Multimodal LLMs in the Lens of Few-Shot Learning
It provides insights for optimizing MLLMs in visual-language tasks, but the work is incremental as it focuses on evaluating existing methods on a known benchmark.
This study evaluated multimodal large language models (MLLMs) on the VALSE benchmark, finding that few-shot in-context learning and chain-of-thought prompting significantly boost performance, with models pretrained on captioning data showing superior zero-shot results and those on interleaved data benefiting more from few-shot learning.
The linguistic capabilities of Multimodal Large Language Models (MLLMs) are critical for their effective application across diverse tasks. This study aims to evaluate the performance of MLLMs on the VALSE benchmark, focusing on the efficacy of few-shot In-Context Learning (ICL), and Chain-of-Thought (CoT) prompting. We conducted a comprehensive assessment of state-of-the-art MLLMs, varying in model size and pretraining datasets. The experimental results reveal that ICL and CoT prompting significantly boost model performance, particularly in tasks requiring complex reasoning and contextual understanding. Models pretrained on captioning datasets show superior zero-shot performance, while those trained on interleaved image-text data benefit from few-shot learning. Our findings provide valuable insights into optimizing MLLMs for better grounding of language in visual contexts, highlighting the importance of the composition of pretraining data and the potential of few-shot learning strategies to improve the reasoning abilities of MLLMs.