Introducing the MMWorld benchmark

[jkooy]

Dear Qwen team,

We are a big fan of your Qwen series and noticed that you have evaluated your models on several video-language benchmarks. We were wondering if you might be interested in evaluating your models on our MMWorld benchmark (https://arxiv.org/abs/2406.08407). MMWorld is designed to assess models' reasoning capabilities across various reasoning tasks and disciplines and could serve as a useful evaluation benchmark for your model development. Thank you!

[ShuaiBai623]

Thank you for your recognition and good video evaluation work. We will test it after the Spring Festival.