Loading results…
Success Rate
Fraction of correct predictions (model × action, N = 250 samples)
Mean Max-Distance of Correct Predictions
Lower is better — average maximum atom displacement (Å) over correct predictions only
Outcome Breakdown per Action
Stacked bars show the fraction of each result type across models
Correct
StructureMismatch
AtomCountMismatch
CIFParsingError
OutputFormatError
Other