Picking more hotspots seems to be more effective for binder generation contrary to Readme

[ackbar03]

Hi,

In the Readme (https://github.com/RosettaCommons/RFdiffusion#picking-hotspots), it suggests 5-6 hotspots because "the model is expecting to have to make more contacts than you specify".

However, during my testing, I found that specifying more hotspots helped (usually more than 10), and within 15 Angstroms instead of 10. This prevents the finder from being generated in completely different location, which sometimes happens when there are not enough hot-spots specified.

Is this expected behavior? Should I be using less hotspots (3-6 as recommended) but modify the weights related to the hotspot potentials instead?

Thanks