Performance Misaligned with your paper.

[guyuchao]

Using the provided example, I can align the performance with your readme. However, the performance in the readme is different with your paper. What causes this difference?

[LeyRio]

Thank you for your interest in our work. There are two reasons why the metrics in the README differ from those in the paper. First, when releasing the open-source version, we removed the "people" category from the benchmark because expressions like 'a red/green/yellow... people' seemed a bit odd. Second, the weights in the open-source version have stronger control capabilities compared to those in the paper, resulting in higher metrics.

[guyuchao]

thanks for reply