Artifact-Bench: Evaluating MLLMs on Detecting and Assessing the Artifacts of AI-Generated Videos

TL;DR

Artifact-Bench is a new benchmark designed to evaluate Multimodal Large Language Models (MLLMs) on their ability to detect, compare, and analyze artifacts in AI-generated videos across various styles.

Contribution

The paper introduces Artifact-Bench, a comprehensive hierarchical taxonomy and evaluation tasks for assessing MLLMs' artifact perception in diverse AI-generated videos.

Findings

Many MLLMs perform near random in artifact detection tasks.

Significant misalignment exists between MLLM judgments and human preferences.

Current MLLMs have limited reliability in evaluating AI-generated video realism.

Abstract

Recent video generative models have greatly improved the realism of AI-generated videos, yet their outputs still exhibit artifacts such as temporal inconsistencies, structural distortions, and semantic incoherence. While Multimodal Large Language Models (MLLMs) show strong visual understanding capabilities, their ability to perceive and reason about such artifacts remains unclear. Existing benchmarks often lack systematic evaluation of artifact-aware perception and fine-grained diagnostic reasoning, especially across diverse AI-generated video domains beyond photorealistic content. To address this gap, we introduce Artifact-Bench, a comprehensive benchmark for evaluating MLLMs on AI-generated video artifact detection and analysis. We first establish a three-level hierarchical taxonomy of realism artifacts, covering photorealistic, animated, and CG-style videos. Based on this taxonomy, Artifact-Bench defines three complementary tasks: real vs. AI-generated video classification, pairwise realism comparison, and fine-grained artifact identification. Experiments on 19 leading MLLMs reveal substantial limitations in artifact perception and reasoning, with many models approaching random or even below-random performance in challenging settings. We further observe significant misalignment between MLLM judgments and human perceptual preferences, highlighting their limited reliability as general evaluators for AI-generated video realism.