What Makes Multimodal In-Context Learning Work?

TL;DR

This paper investigates the effectiveness of multimodal in-context learning in large models, revealing that it mainly depends on text and is often no better than simple voting strategies, highlighting biases and limitations.

Contribution

It provides a comprehensive analysis of multimodal in-context learning, showing its reliance on text and comparing advanced strategies with simple voting, which is a novel insight.

Findings

M-ICL mainly relies on text-driven mechanisms.

Advanced-ICL strategies do not outperform majority voting.

Identifies biases and limitations in M-ICL.

Abstract

Large Language Models have demonstrated remarkable performance across various tasks, exhibiting the capacity to swiftly acquire new skills, such as through In-Context Learning (ICL) with minimal demonstration examples. In this work, we present a comprehensive framework for investigating Multimodal ICL (M-ICL) in the context of Large Multimodal Models. We consider the best open-source multimodal models (e.g., IDEFICS, OpenFlamingo) and a wide range of multimodal tasks. Our study unveils several noteworthy findings: (1) M-ICL primarily relies on text-driven mechanisms, showing little to no influence from the image modality. (2) When used with advanced-ICL strategy (like RICES), M-ICL is not better than a simple strategy based on majority voting over context examples. Moreover, we identify several biases and limitations of M-ICL that warrant consideration prior to deployment. Code available at https://gitlab.com/folbaeni/multimodal-icl