mTVR uses the exact same evaluation script as TVR evaluation, please refer to details there. Specifically, you should be able to find the following resources:
- Task Definition
- How to construct a prediction file?
- Run Evaluation
To test your model's performance on test-public set, please submit both val and test-public predictions to our Codalab evaluation server. The submission file should be a single .zip file (no enclosing folder) that contains the 4 prediction files zh_tvr_val_submission.json, zh_tvr_test_public_submission.json, en_tvr_val_submission.json and en_tvr_test_public_submission.json. Each of the *submission.json file should be formatted as instructed in TVR Evaluation.