Vlm-Long-Doc-Qa-Single-Page-Qa

_{Made with ❤️ using 🎨 NeMo Data Designer}

---

单页问答数据集，每条样本基于 PDF 单页图像生成，包含问题、答案、思维链推理过程及质量评分。问题类型涵盖数值计算、列表提取、布局定位等。

---

🚀 Quick Start

from datasets import load_dataset

# Load the main dataset
dataset = load_dataset("lieeli/vlm-long-doc-qa-single-page-qa", "data", split="train")
df = dataset.to_pandas()

---

📊 Dataset Summary

• 📈 Records: 999

• 📋 Columns: 10

• ✅ Completion: 99.9% (1,000 requested)

---

📋 Schema & Statistics

Column	Type	Column Type	Unique (%)	Null (%)	Details
question_type	string	sampler	9 (0.9%)	0 (0.0%)	category
question	string	llm-text	997 (99.8%)	0 (0.0%)	Tokens: 46 out / 3495 in
answer	string	llm-text	625 (62.6%)	0 (0.0%)	Tokens: 9 out / 1482 in
quality_score	string	llm-text	3 (0.3%)	0 (0.0%)	Tokens: 1 out / 1513 in

---

⚙️ Generation Details

Generated with 10 column configuration(s):

• llm-text: 3 column(s)

• sampler: 1 column(s)

• seed-dataset: 6 column(s)

📄 Full configuration available in builder_config.json and detailed metadata in metadata.json.

---

📚 Citation

If you use Data Designer in your work, please cite the project as follows:

@misc{nemo-data-designer,
  author = {The NeMo Data Designer Team, NVIDIA},
  title = {NeMo Data Designer: A framework for generating synthetic data from scratch or based on your own seed data},
  howpublished = {\url{https://github.com/NVIDIA-NeMo/DataDesigner}},
  year = 2026,
  note = {GitHub Repository},
}

---

💡 About NeMo Data Designer

NeMo Data Designer is a general framework for generating high-quality synthetic data that goes beyond simple LLM prompting. It provides:

• Diverse data generation using statistical samplers, LLMs, or existing seed datasets
• Relationship control between fields with dependency-aware generation
• Quality validation with built-in Python, SQL, and custom local and remote validators
• LLM-as-a-judge scoring for quality assessment
• Fast iteration with preview mode before full-scale generation

For more information, visit: https://github.com/NVIDIA-NeMo/DataDesigner (pip install data-designer)