MemeArena: Automating Context-Aware Unbiased Evaluation of Harmfulness Understanding for Multimodal Large Language Models

TL;DR

MemeArena introduces a novel, context-aware, and unbiased evaluation framework for assessing multimodal large language models' understanding of harmful content in memes, addressing limitations of existing binary classification methods.

Contribution

It presents an agent-based arena-style framework that simulates diverse contexts and integrates multiple viewpoints for fairer, more nuanced evaluation of mLLMs' harmfulness comprehension.

Findings

Reduces evaluation bias effectively

Aligns judgment results with human preferences

Provides comprehensive insights into mLLMs' interpretive abilities

Abstract

The proliferation of memes on social media necessitates the capabilities of multimodal Large Language Models (mLLMs) to effectively understand multimodal harmfulness. Existing evaluation approaches predominantly focus on mLLMs' detection accuracy for binary classification tasks, which often fail to reflect the in-depth interpretive nuance of harmfulness across diverse contexts. In this paper, we propose MemeArena, an agent-based arena-style evaluation framework that provides a context-aware and unbiased assessment for mLLMs' understanding of multimodal harmfulness. Specifically, MemeArena simulates diverse interpretive contexts to formulate evaluation tasks that elicit perspective-specific analyses from mLLMs. By integrating varied viewpoints and reaching consensus among evaluators, it enables fair and unbiased comparisons of mLLMs' abilities to interpret multimodal harmfulness. Extensive experiments demonstrate that our framework effectively reduces the evaluation biases of judge agents, with judgment results closely aligning with human preferences, offering valuable insights into reliable and comprehensive mLLM evaluations in multimodal harmfulness understanding. Our code and data are publicly available at https://github.com/Lbotirx/MemeArena.