Update README.md

README.md

@@ -16,14 +16,20 @@ tags:
 - mgt-detection
 - ai-detection
 metrics:
-- accuracy
-- precision
-- recall
-- f1
+- type: accuracy
+  value: 0.973
+- type: precision
+  value: 1.000
+- type: recall
+  value: 0.945
+- type: f1
+  value: 0.972
 ---
 
 The hosted inference here, for some reason, does not work. It seems like this is because the hosted inference ignores the attention-mask when doing predictions. It works with pipelines however, and when including attention-mask in the model, e.g. `**encodings`.
 
-[will update readme later]
+This is a text classification for detecting machine-generated text based on the bloomz-560m by BigScience (see https://huggingface.co/bigscience/bloomz-560m).
+The model is fine-tuned on generations by GPT-2.
 
-This is a text classification for detecting machine-generated text based on the {_name_or_path}
+**NOTE**: the hosted inference does not work. To do inference, please download the model to perform inference.
+**NOTE**: when not using the attention-head in the predictions, the predictions will be wrong.