Viettel AI Challenge 2025 - QA Reasoning Dataset
Dataset Description
Fine-tuning dataset for Qwen3 models, generated from RAG-based question answering on Vietnamese technical documents. Contains reasoning-style responses with chain-of-thought evaluation of multiple-choice questions.
Dataset Structure
Format: Conversational (Qwen3 chat template)
Splits: train (80%), validation (20%)
Total Examples: 307
Language: Vietnamese + English (mixed)
Data Fields… See the full description on the dataset page: https://huggingface.co/datasets/tisu1902/var-public-test.
