# VGI-Bench：用 27 项任务探测视频生成模型的视觉智能

- 来源：HuggingFace Daily Papers（社区热门论文）
- 发布时间：2026-08-20 08:00
- AIHOT 分数：51
- AIHOT 链接：https://aihot.virxact.com/items/cmtaxzs6k0d35roamsmm2gh4z
- 原文链接：https://arxiv.org/abs/2608.19583

## AI 摘要

VGI-bench 提出一套视频生成模型视觉推理评测基准，含 27 个任务、810 个实例，按任务领域和技能标签两级分类。评测显示当前最强模型 Seedance 2.0 仅达 51.0% 准确率，生成系统虽能解决部分视觉推理任务但远不可靠。分析还揭示了输出失败模式、输入条件敏感性和去噪阶段自我修正有限等问题，代码与数据将开源。

## 正文

Recent studies suggest that video generation models can exhibit certain forms of zero-shot visual reasoning through generated frames. Yet reliable evaluation remains challenging: benchmarks should adopt inputs aligned with the visual priors of current video models, require valid evolving processes rather than only plausible final states, and calibrate task difficulty to remain challenging yet partly feasible. To this end, we introduce VGI-bench, containing 27 tasks and 810 instances, organized by a two-level taxonomy of task domains and skill tags for fine-grained evaluation of visual reasoning capabilities of video generation models. Our evaluations show that current generative systems can solve a subset of visually grounded reasoning tasks, but remain far from reliable, with even the strongest model, Seedance~2.0, achieving only 51.0% under our evaluation criteria. Our analysis further explore the output failure modes, input condition sensitivity, performance transfer boundary from synthetic fine-tuning, and internal denoising perspective revealing limited self-correction, where later steps mainly refine early hypotheses rather than correct reasoning errors. We hope VGI-bench will help stimulate the development of next-generation video generation models. We will release our code and data.