Paper question

[Serega6678]

Hello, thank you for your work

Can you share the results of the impact of each of the training steps you've made (subgoal guidance & error recovery vs just 1-stage fine-tunning)

Thank you

[Sirius-L1]

Hi @Serega6678,

Thank you for your interest in our work.

Our research is currently a work in progress, and we are in the process of conducting comprehensive and scientific experiments. This includes a detailed analysis of the individual contributions of each training stage to the overall model performance.

Preliminary results suggest that the second stage contributes significantly more to the performance improvement than the initial fine-tuning stage. We are actively investigating the underlying reasons for this observation.

We plan to update our paper and repository with the complete experimental results and ablation studies as soon as they are finalized. We will be sure to notify you when these updates are available.

Best regards,

The InfiGUI-R1 Team