Fine-tuning with self-correction

[HaochenZ11]

Hi,

Thanks for the great work! I noticed in your main figure and method section of the paper the fine-tuning with self-correction is a component of your model. However, I noticed you had an additional table with the effect of fine-tuning on a subset of Nr3D separate from the main results table.

I just wanted to clarify, for your main results table (Table 1) on Nr3D/Sr3D results, did your best-performing method with GPT-4-P make use of this fine-tuning or not?