---
language:
  - zh
tags:
  - mbart-50
license: apache-2.0
datasets:
  - CATSLU
metrics:
  - accuracy
  - f1
---

This model is `mbart-large-50-many-to-many-mmt` model fine-tuned on the text part of [CATSLU](https://sites.google.com/view/CATSLU/home) spoken language understanding dataset.

The average scores on the four test sets are 82.56% and 73.48% for accuracy and F1 respectively.