Spaces:

---

PrunaAI

/

P-Bench

Like 25

Running

Hiii Minette! Amazing job overall, I just have a few comments. Again writing them all in batches here!

• In the leaderboard section, we named the image to video dataset as "Image-to-Video-Benchmark" it should be Vbench-i2V.
• I noticed in the About section we have some missing information:
  1 For VBench - 2.0 Dataset We should say that we have 100 samples. The categories were weighted proportional to the amount of prompts the original dataset has.
  2 For VBench - I2V We used the "subject" attribute of the images to weight the amount of prompts again proportional to the amount of those subjects appearing in the original dataset.
• In the About tab, under setup, for generation (text-to-video) and (image-to-video) we mention something about the cost mode. I am a little confused by this. Maybe here we can just mention that we have 100 samples and each video generated is 5 seconds for both of these modalities?
  Thank you very much!!