UVE: Are MLLMs Unified Evaluators for AI-Generated Videos?
Yuanxin Liu, Rui Zhu, Shuhuai Ren, Jiacong Wang, Haoyuan Guo, Xu Sun, Lu Jiang
2025-03-21
Summary
This paper explores whether AI models that understand both images and language can be used to automatically judge the quality of videos made by AI.
What's the problem?
It's hard to know if AI-generated videos are good because current methods for evaluating them are limited and don't cover all the important aspects.
What's the solution?
The researchers tested whether AI models could be used as a single, all-purpose tool to evaluate different aspects of AI-generated videos.
Why it matters?
This work matters because it could lead to better ways to automatically assess and improve the quality of AI-generated videos.
Abstract
With the rapid growth of video generative models (VGMs), it is essential to develop reliable and comprehensive automatic metrics for AI-generated videos (AIGVs). Existing methods either use off-the-shelf models optimized for other tasks or rely on human assessment data to train specialized evaluators. These approaches are constrained to specific evaluation aspects and are difficult to scale with the increasing demands for finer-grained and more comprehensive evaluations. To address this issue, this work investigates the feasibility of using multimodal large language models (MLLMs) as a unified evaluator for AIGVs, leveraging their strong visual perception and language understanding capabilities. To evaluate the performance of automatic metrics in unified AIGV evaluation, we introduce a benchmark called UVE-Bench. UVE-Bench collects videos generated by state-of-the-art VGMs and provides pairwise human preference annotations across 15 evaluation aspects. Using UVE-Bench, we extensively evaluate 16 MLLMs. Our empirical results suggest that while advanced MLLMs (e.g., Qwen2VL-72B and InternVL2.5-78B) still lag behind human evaluators, they demonstrate promising ability in unified AIGV evaluation, significantly surpassing existing specialized evaluation methods. Additionally, we conduct an in-depth analysis of key design choices that impact the performance of MLLM-driven evaluators, offering valuable insights for future research on AIGV evaluation. The code is available at https://github.com/bytedance/UVE.