Update README.md

README.md

@@ -62,10 +62,6 @@ print(f"Standard: {normalize_text(dialect_text)}")
 
 If you encounter language-token issues, set `tokenizer.src_lang = "vi_VN"` and pass `forced_bos_token_id=tokenizer.lang_code_to_id["vi_VN"]` to `model.generate()`.
 
-## Reference results
-
-The paper's mBART-large-50 baseline (trained on the `train` split only) achieves **BLEU 0.8166 · ROUGE-L 0.9384 · METEOR 0.8925** on ViDia2Std. This full-corpus model cannot be legitimately evaluated on that test set.
-
 ## Citation
 
 ```bibtex