Question on the OC20NEB dataset

[anyangml]

Hi @brookwander,

I recently read your paper—really fascinating work! I’m interested in testing some MLIP models with the dataset, but I couldn’t find the DFT labels for the initial and final frames. It looks like all intermediate steps are labeled, but without those values, I’m unsure how to compute the ground truth energy barrier.

Am I missing something? I’d appreciate any guidance on how to determine it.

Thanks in advance!