Decoder benchmarking and auto-reconstruction error

[OliviaViessmann]

Hello,

thank you for providing details on the model in the paper's appendix.
I am trying to get an idea on the backbone and side-chain rmse values on the validation datasets of CAMEO, CASP14 and 15 on page 32 of the supplementary. They are not numerically mentioned in the text, so I am trying to guesstimate them by eye from the figure S3.
Could you provide the actual values? That would be tremendously helpful for benchmarking and fidelity

Thanks
Olivia