Order Matters: Exploring Order Sensitivity in Multimodal Large Language Models

TL;DR

This paper investigates how input order affects Multimodal Large Language Models' performance, revealing order sensitivity and proposing strategies to improve accuracy by leveraging attention patterns and a new evaluation metric.

Contribution

It uncovers order sensitivity in MLLMs, demonstrates how input positioning impacts performance, and introduces Position-Invariant Accuracy to better evaluate models.

Findings

Order changes cause performance fluctuations in MLLMs.

Positioning key content at attention hotspots improves results.

Proposed PIA metric reduces order bias in evaluation.

Abstract

Multimodal Large Language Models (MLLMs) utilize multimodal contexts consisting of text, images, or videos to solve various multimodal tasks. However, we find that changing the order of multimodal input can cause the model's performance to fluctuate between advanced performance and random guessing. This phenomenon exists in both single-modality (text-only or image-only) and mixed-modality (image-text-pair) contexts. Furthermore, we demonstrate that popular MLLMs pay special attention to certain multimodal context positions, particularly the beginning and end. Leveraging this special attention, we place key video frames and important image/text content in special positions within the context and submit them to the MLLM for inference. This method results in average performance gains of 14.7% for video-caption matching and 17.8% for visual question answering tasks. Additionally, we propose a new metric, Position-Invariant Accuracy (PIA), to address order bias in MLLM evaluation. Our research findings contribute to a better understanding of Multi-Modal In-Context Learning (MMICL) and provide practical strategies for enhancing MLLM performance without increasing computational costs.