Vietnamese fact-checking verifier (XLM-R base)

Input order is (evidence, claim). Labels: SUPPORTED, REFUTED, and NOT_ENOUGH_INFO. Training/validation/test are split by source document.

Validation Macro-F1: 0.9665.

This model was trained on synthetic/silver claims and must not be treated as a human-validated factuality oracle.

Downloads last month
-
Safetensors
Model size
0.3B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Loctran123/vietnamese-fact-checking-verifier-xlm-roberta-base

Finetuned
(4172)
this model