datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
dataset_vietnameseVietnamese-OpenO1-SFTOriginal dataset: https://huggingface.co/datasets/qingy2024/OpenO1-SFT-Cleaned
This dataset is a Vietnamese translated version of qingy2024/OpenO1-SFT-Cleaned. Please cite the original dataset if you find it useful in your work.
Translated to Vietnamese with context-aware using gemini-flash-2.0-exp via this repo:
https://github.com/vTuanpham/Large_dataset_translator.
Roughly 4 hours for 500 examples.
Vietnamese-lmms-lab-LLaVA-Video-178K-gg-translated
Dataset Card for 5CD-AI/Vietnamese-lmms-lab-LLaVA-Video-178K-gg-translated
This translated dataset includes:
LLaVA-Video-178K: 178,509 caption entries, 960,791 open-ended QA (question and answer) items, and 196,198 multiple-choice QA items.
The video source of the original dataset is in this repo: lmms-lab/LLaVA-Video-178K
Vietnamese-Multi-turn-Chat-AlpacaVietnamese-yfcc15m-OpenAICLIPKTO-mix-14k-vietnamese-groqOriginal dataset: https://huggingface.co/datasets/trl-lib/kto-mix-14k
This dataset is a KTO-formatted version of argilla/dpo-mix-7k. Please cite the original dataset if you find it useful in your work.
Translated to Vietnamese with context-aware using Groq Llama3.3 70B* via this repo:
https://github.com/vTuanpham/Large_dataset_translator.
Roughly 9 hours for 2k examples.
Usage
from datasets import load_dataset
kto_mix_14k_vi =… See the full description on the dataset page: https://huggingface.co/datasets/1TuanPham/KTO-mix-14k-vietnamese-groq.Vietnamese-Legal-QA
Vietnamese Legal QA — Question Specificity
Phân loại độ cụ thể của câu hỏi pháp luật dân sự Việt Nam: broad (hỏi khái
quát, phải tổng hợp nhiều chế định) hay narrow (hỏi vào một tình huống / một
điều luật xác định). Dùng để định tuyến truy vấn trong hệ RAG pháp luật.
Cấu trúc
Mỗi dòng là một câu hỏi kèm vết gán nhãn. Hai dòng cùng pair_id là một cặp
đối chứng sinh từ cùng một điều luật — một broad, một narrow.
Trường
Ý nghĩa
item_id, pair_id… See the full description on the dataset page: https://huggingface.co/datasets/ThanhVu101/Vietnamese-Legal-QA.Vietnamese-magpie-ultra-v0.1Original dataset: https://huggingface.co/datasets/argilla/magpie-ultra-v0.1
### Dataset Summary
`magpie-ultra` it's a synthetically generated dataset for supervised fine-tuning using the new Llama 3.1 405B-Instruct model, together with other Llama models like `Llama-Guard-3-8B` and `Meta-Llama-3.1-8B-Instruct`.
The dataset contains challenging instructions and responses for a wide variety of tasks, such as Coding & debugging, Math, Data analysis, Creative Writing, advice seeking, or… See the full description on the dataset page: https://huggingface.co/datasets/1TuanPham/Vietnamese-magpie-ultra-v0.1.Vietnamese-alpaca-gpt4-gg-translatedVietnamese_RAG
Dataset Card for Dataset Name
Vi's RAG is an comprehensive Vietnamese dataset optimized for RAG Evaluation, build by ZD AI lab and release under Apache license 2.0.
Dataset Details
There are four datasets in this card :
Vietnamese version of Expert QA that we utilize the strong translation ability of GPT-4 for translation task
RAG ViQuAD which was carefully chosen from UIT-ViQuAD2.0 with additional context column filtered by title
Legal RAG and BKAI_RAG are long form RAG… See the full description on the dataset page: https://huggingface.co/datasets/sailor2/Vietnamese_RAG.Vietnamese-Salesforce-xlam-function-calling-60k-gg-translatedvietnamese-traffic-sign-vqa
Vietnamese Traffic Sign VQA
Visual Question Answering dataset for Vietnamese traffic signs.
Built from Kaggle VNTS (CC BY-SA 4.0).
Statistics
Split
Images
QA Pairs
QA/Image
Train
2,193
104,146
47.5
Val
272
12,944
47.6
Test
271
12,966
47.8
Total
2,736
130,056
47.5
Question Types
12 types: yes_no, count, sign_type, color, shape, location, attribute, negative, spatial_rel, count_total, multi_object, context
Format
{… See the full description on the dataset page: https://huggingface.co/datasets/Anakonkai/vietnamese-traffic-sign-vqa.Vietnamese-nampdn-ai-tiny-webtext-gg-translatedVietnamese-395k-meta-math-MetaMathQA-gg-translatedVietnamese-SFT-Corpus-V2
🇻🇳 Vietnamese SFT Corpus V2.1 (Balanced Safety & Anti-Over-Refusal)
Vietnamese SFT Corpus V2.1 là tập dữ liệu Tinh chỉnh có Giám sát (Supervised Fine-Tuning - SFT) chuẩn công nghiệp dành cho mô hình ngôn ngữ lớn (LLM) tiếng Việt. Tập dữ liệu được thiết kế nhằm phục vụ huấn luyện trợ lý ảo thông minh, hội thoại tự nhiên, suy luận logic, đồng thời đặc trị triệt để hiện tượng "từ chối lười biếng / từ chối nhầm" (Lazy Refusal / Over-Refusal) vốn xuất hiện phổ biến ở các mô… See the full description on the dataset page: https://huggingface.co/datasets/Phuc-HugigFace/Vietnamese-SFT-Corpus-V2.Vietnamese-ShareGPT4Vision-gg-translatedchinese_vietnamese_bilingual_wangwen本数据集是一个中文到越南语的机器翻译数据集。数据集构造自较为受欢迎的网络小说，首先从越南语的小说站点根据排行榜看有哪些书比较受欢迎，看看哪些书是从对应的中文网文小说翻译而来的（大部分都是）。
拿到同一本书的中文版本和越南语版本后，就可以进行alignment。如果翻译是忠于原著的，那么
每个章节都能对上
同一个章节中的每个句子都能对上
实操过程中作者踩了很多坑，比如
作者的写作习惯不一样，无法把txt文本有效地切割成chapters
中文和越文版本的小说正文中有可能夹杂一些广告，要尽量过滤掉这些噪音
长篇网文有2000多chapter，中文版本和越文版本都可能丢失一些章节，要过滤掉无法对齐的章节
对齐算法是作者自己设计的，参考了transportation theory，以章节对齐为例。首先计算中文章节和越文章节两两之间的相似度，然后由动态规划算法寻找一条最优路径，给每一个中文章节asign一个越文章节。大致的过程如下图所示：
于是相似度计算就是其中的关键因素，对齐章节和对齐章节里的句子采取不同的相似度matric。
Chapter… See the full description on the dataset page: https://huggingface.co/datasets/kaihe/chinese_vietnamese_bilingual_wangwen.vietnamese-financial-summary
Vietnamese Financial News Summarization with Number Preservation
Vietnamese-liuhaotian-llava_v1_5_mix665k-gg-translatedVietnamese-Intel-orca_dpo_pairs-gg-translatedvietnamese-sft-10k
Vietnamese Instruction-Following Dataset (10K)
This dataset comprises 10,000 Vietnamese instruction-style prompt-response pairs curated for supervised fine-tuning (SFT) of language models. It aims to improve conversational and instruction-following abilities in the Vietnamese language, with coverage across diverse social, cultural, and emotional contexts.
Format: JSONL (one object per line)
Fields: "prompt" (instruction or user message), "response" (assistant reply)
Language:… See the full description on the dataset page: https://huggingface.co/datasets/vlinhd11/vietnamese-sft-10k.vietnamese-ms-hs-textbook-math-300K
Vietnam Math Textbook Dataset
📘 Introduction
The Vietnam Math Textbook Dataset is a comprehensive resource for mathematical knowledge, directly extracted and compiled from standard Vietnamese math textbooks.Its goal is to provide a rich repository of information—including concepts, exercises, detailed solutions, and explanations in Vietnamese—to support learning, research, and the development of educational applications.
📂 Key Contents
The dataset… See the full description on the dataset page: https://huggingface.co/datasets/LuminarAI/vietnamese-ms-hs-textbook-math-300K.vietnamese-dataset
Vietnamese Dialogue Dataset
Dữ liệu hội thoại tiếng Việt ngắn gọn, tự nhiên, dùng để fine-tune các mô hình ngôn ngữ như Mamba, LLaMA, Gemma. Nội dung bao gồm giao tiếp hàng ngày, câu hỏi thường gặp, phản hồi cảm xúc...
✅ Chuẩn định dạng JSONL
✅ Sẵn sàng dùng cho huấn luyện instruction tuning
✅ Không chứa nội dung nhạy cảm hay vi phạm
📌 Tạo bởi @hoanghai2110 để phục vụ cộng đồng mã nguồn mở AI Việt Nam.
vietnamese-evidence-retrieval-indexes
Vietnamese Evidence Retrieval Indexes
Prebuilt exact dense and sparse indexes for
Loctran123/vietnamese-evidence-corpus-embeddings-e5-large at revision e928944361ca7d4c80f80d19bec52ebad55a4f7f.
Rows: 52,605
Source embedding shards: 11
Dense: FAISS IndexFlatIP, 1024 dimensions
Sparse: BM25S Lucene BM25 (k1=1.5, b=0.75)
BM25 content: title repeated 2 times + chunk text
Vietnamese tokenization: Unicode word tokens, no stemming and no stopword removal
row_id in metadata.parquet is… See the full description on the dataset page: https://huggingface.co/datasets/Loctran123/vietnamese-evidence-retrieval-indexes.vietnamese-evidence-retrieval-indexes-v2-r1
Vietnamese Evidence Retrieval Indexes
Prebuilt exact dense and sparse indexes for
Loctran123/vietnamese-evidence-corpus-embeddings-e5-large-v2-r1 at revision 2a18d35b6ea2e078db95c1aacdc2a28947268b4e.
Rows: 63,699
Source embedding shards: 13
Dense: FAISS IndexFlatIP, 1024 dimensions
Sparse: BM25S Lucene BM25 (k1=1.5, b=0.75)
BM25 content: title repeated 2 times + chunk text
Dense input: title + text
Dense rows: deduplicated by content hash
Vietnamese tokenization: Unicode word… See the full description on the dataset page: https://huggingface.co/datasets/Loctran123/vietnamese-evidence-retrieval-indexes-v2-r1.vietnamese-dpo-10k
Vietnamese DPO Dataset (10K)
This dataset contains 10,000 Vietnamese prompt-response pairs in the Direct Preference Optimization (DPO) format, including a "prompt", a "chosen" response (preferred), and a "rejected" response (less preferred or misaligned). It is intended for training language models to better align with human-preferred responses, particularly in edge cases involving social sensitivity, rudeness, or toxicity.
Format: JSONL (one object per line)
Fields:
"prompt":… See the full description on the dataset page: https://huggingface.co/datasets/vlinhd11/vietnamese-dpo-10k.Vietnamese-o1-journeyOriginal dataset: https://huggingface.co/datasets/GAIR/o1-journey
This dataset is a Vietnamese translated version of GAIR/o1-journey. Please cite the original dataset if you find it useful in your work.
Translated to Vietnamese with context-aware using gemini-flash-2.0-exp via this repo:
https://github.com/vTuanpham/Large_dataset_translator.
Roughly 2 hours for 649 examples.
Vietnamese-meta-math-MetaMathQA-40K-gg-translatedVietnamese-microsoft-orca-math-word-problems-200k-gg-translatedVietnamese-cosmos-qa-gg-translated