The score in the paper is confusing

[saikouMonika]

Thank you very much for your work. However, I have a few minor questions.

Firstly, the fine-tuned base model in your repository is Show-o-512x512, which is supposed to have a score of 0.68 on GenEval, but why the baseline score you listed is 0.53?

Additionally, there appears to be an inconsistency in your multi-model understanding scores. You directly used Show-o's scores, but for POPE metrics you employed the VQGAN-based version of Show-o, while other scores were derived from the CLIP-based version of Show-o. This discrepancy is quite confusing.

Could you please open-source the evaluation code to help verify these results?

[Cominclip]

Thank you for your attention to our work.

We sincerely apologize for any confusion caused. The baseline score of GenEval was extracted from multiple papers (Janus, Vila-u, Emu3). Regarding other metrics you mentioned, such as POPE, we will conduct a reevaluation and review of the model(including GenEval), and will subsequently release the complete evaluation code or methodology.