Identical Results to AgentCPM-GUI

[stellardust66]

Hi team, first of all, great work on the project! I was reading the paper and had some questions:

1. Results in Table 7.1 and 7.2: I noticed that the results reported in these tables for your model appear to be identical to the results listed for AgentCPM-GUI Grounding Benchmark and AgentCPM-GUI Agent Benchmark.
   I wanted to confirm if this is intended (e.g., perhaps due to the experimental setup or a shared baseline component?) or if there might have been a typo.

Your results:

Their results:

2. Could you inform us about any planned release date for the code?

Thank you very much!

[qianc62]

Hello stellardust66,

Regarding the tables in the paper, some errors inevitably occurred during version iterations. We have promptly identified and submitted a new version to arXiv. Thank you for your attention and reminder! Furthermore, to ensure reproducibility, the relevant models and code for this work have now been open-sourced. We sincerely invite you to review or run the inference and evaluation yourself. If you have any further questions, we will continue to answer them. Thank you again!
Paper link: https://arxiv.org/pdf/2509.02444
Open-source model: https://huggingface.co/ffcosmos/AppCopilot/tree/main
User manual: https://github.com/OpenBMB/AppCopilot