FakeBench: Probing Explainable Fake Image Detection via Large Multimodal Models

TL;DR

This paper introduces FakeBench, a multimodal database and evaluation framework leveraging large multimodal models for explainable fake image detection, emphasizing transparency and forensic interpretability over traditional binary classification.

Contribution

It pioneers the use of large multimodal models for explainable fake image detection and provides a new multimodal dataset with human-like forgery descriptions for comprehensive evaluation.

Findings

LMMs show strengths in detection and reasoning tasks.

Evaluation reveals specific merits and weaknesses of different LMMs.

FakeBench promotes transparency and forensic interpretability in fake image detection.

Abstract

The ability to distinguish whether an image is generated by artificial intelligence (AI) is a crucial ingredient in human intelligence, usually accompanied by a complex and dialectical forensic and reasoning process. However, current fake image detection models and databases focus on binary classification without understandable explanations for the general populace. This weakens the credibility of authenticity judgment and may conceal potential model biases. Meanwhile, large multimodal models (LMMs) have exhibited immense visual-text capabilities on various tasks, bringing the potential for explainable fake image detection. Therefore, we pioneer the probe of LMMs for explainable fake image detection by presenting a multimodal database encompassing textual authenticity descriptions, the FakeBench. For construction, we first introduce a fine-grained taxonomy of generative visual forgery concerning human perception, based on which we collect forgery descriptions in human natural language with a human-in-the-loop strategy. FakeBench examines LMMs with four evaluation criteria: detection, reasoning, interpretation and fine-grained forgery analysis, to obtain deeper insights into image authenticity-relevant capabilities. Experiments on various LMMs confirm their merits and demerits in different aspects of fake image detection tasks. This research presents a paradigm shift towards transparency for the fake image detection area and reveals the need for greater emphasis on forensic elements in visual-language research and AI risk control. FakeBench will be available at https://github.com/Yixuan423/FakeBench.