Fetching the paper…

Beyond calibration: estimating the grouping loss of modern neural networks · Around