Dataset · airesearch
Culturally-aware Thai instruction dataset
airesearch/WangchanThaiInstruct
Human-written Thai instruction dataset for training and evaluating models with local cultural awareness.
What's inside
Contains human-written Thai instruction texts (volume of 10 to 100 thousand examples). The data covers 4 subject areas and 7 task types, including text generation, question answering, summarization, and classification. Presented at EMNLP'25 and incorporates cultural features.
What you can do with it
- SFT fine-tuning of LLMs for Thai language processing
- Evaluating instruction-following quality in a low-resource language environment
- Training multi-task models for Thai classification and summarization
- Tasks
- Text generation, Question answering, Summarization, Text classification
- Languages
- Thai
- Size
- 10K–100K rows
- Downloads this month
- 686
- Likes
- 33
Models trained on it
- mradermacher/Qwen2.5-7B-ThaiInstruct-i1-GGUF 923
- mradermacher/Llama-3.1-8B-ThaiInstruct-i1-GGUF 715
- mradermacher/Gemma2-9b-WangchanLIONv2-instruct-i1-GGUF 613
- Eggchad/mango-artifactsText generation 593
- mradermacher/Qwen2.5-7B-ThaiInstruct-GGUF 579
- mradermacher/Gemma2-9b-WangchanLIONv2-instruct-GGUF 262
- mradermacher/Llama-3.1-8B-ThaiInstruct-GGUF 199
- aisingapore/Gemma2-9b-WangchanLIONv2-instructText generation 79
How to load it
from datasets import load_dataset
ds = load_dataset("airesearch/WangchanThaiInstruct")