Are there any reports on the VBench (or VBench2.0) Total Score of Wan2.2 5B?
I sampled the VBench videos (with gpt enhanced prompts and @704x1280 resolution) and landed on 83.1, which is lower than the scores I get from Wan2.1 1.3B (@480x832), with similar prompts and sampling pipeline. Is it expected?