update.

README.md

@@ -13,9 +13,9 @@ This paper proposes a procedure to synthetically generate a QA dataset, TeaBReaC
 
 We release the following models:
 
-- **A:** Base Models finetuned on target datasets: `{target_dataset}-{base_model}`
+- **A:** Base Models finetuned on target datasets: `{base_model}-{target_dataset}`
 - **B:** Base models pretrained on TeaBReaC: `teabreac-{base_model}`
-- **C:** Base models pretrained on TeaBReaC and then finetuned on target datasets: `teabreac-{target_dataset}-{base_model}`
+- **C:** Base models pretrained on TeaBReaC and then finetuned on target datasets: `teabreac-{base_model}-{target_dataset}`
 
 The `base_model` above can be from: `bart-large`, `t5-large`, `t5-3b`, `nt5-small`, `preasm-large`.
 The `target_dataset` above can be from: `drop`, `tatqa`, `iirc-gold`, `iirc-retrieved`, `numglue`.
@@ -27,7 +27,6 @@ The **A** models are only released for completeness / reproducibility. In your e
 Please checkout the details in our [github repository](https://github.com/stonybrooknlp/teabreac), but in a nutshell:
 
 ```python
-UPDATE_DISCLAIMER
 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 from digit_tokenization import enable_digit_tokenization # digit_tokenization.py from https://github.com/stonybrooknlp/teabreac