BaoNhan/vi-qa-mt5-base-ctx

mT5-base__ctx

Fine-tuned from google/mt5-base for Vietnamese question answering on BaoNhan/FAIR_Dataset_QA.

• Input format: context + cau hoi (with-context)
• Preprocess mode: raw

Metrics (test set)

Metric	Value
Exact_Match	30.45
Token_F1	37.68
ROUGE_1_F1	37.68
ROUGE_2_F1	32.79
ROUGE_L_F1	36.69
BLEU	4.68
chrF	17.94
BERTScore_F1	86.65

Inference

predict.py is included in this repo:

python predict.py --model BaoNhan/vi-qa-mt5-base-ctx --mode raw --use_context --context "..." \
    --question "Viet Nam co bao nhieu tinh thanh?"

Citation

A paper describing this benchmark is in preparation. If you use this code or the released checkpoints, please cite it once it is available — the entry below is a placeholder and will be updated with the final venue, authors, and year.

@misc{TODO_citation_key,
  title        = {TODO: paper title},
  author       = {TODO: author list},
  year         = {TODO},
  howpublished = {TODO: venue / arXiv preprint},
  note         = {Preprint in preparation},
  url          = {https://github.com/chaubaonhan/QA_Huit}
}