https://huggingface.co/datasets/glue includes: * datasets where we may have several input columns ( https://huggingface.co/datasets/glue/viewer/mnli/train , https://huggingface.co/datasets/paws ) * datasets where the labels are floats on a given range, task is regression ( https://huggingface.co/datasets/glue/viewer/stsb/train ) It would be nice to support those in TextClassificationEvaluator.