DATA AUGMENTATION
Collection
High Quality Question Pairs
• 2 items • Updated • 3
Dataset Viewer
The dataset viewer is not available for this dataset.
Unexpected token '<', "<html>
<h"... is not valid JSON
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
❤️ Support Our Mission
🙏 Drop a Heart ❤ For Our Hard Work!
If you believe in the vision of Sovereign Indian AI, please show your support by dropping a heart below. Your encouragement fuels our journey!
Kindly follow us for more updates and contribute to our open-source journey!
SKT DATA AUGMENTATION SUITE
Q-AUGMENTED
SUPERCHARGE YOUR LLM TRAINING
High-Quality Question Pairs • Synthetic Augmentation • Evaluation Ready
A premium collection of augmented question pairs designed to enhance model robustness. Perfect for SFT, RLHF, and rigorous benchmarking without answer bias.
🔄 DATA AUGMENTATION ❓ QUESTION PAIRS 🧠 REASONING BOOST ⚡ UNBIASED EVAL
Dataset Overview
Q-Augmented provides a diverse set of high-quality question pairs generated via advanced augmentation techniques. Unlike standard datasets, this collection focuses on input diversity to help models generalize better across different phrasing styles and complexities.
✨ Why Use Q-Augmented?
- Robustness Training: Expose your model to varied question structures for the same underlying intent.
- Evaluation Benchmark: Test if your model truly understands meaning or just memorizes patterns.
- No Answer Leakage: Pure input pairs allow you to generate fresh answers with your own system prompts.
- SFT & RLHF Ready: Ideal base for creating preference pairs or expanding instruction datasets.
AUGMENT YOUR INTELLIGENCE
Better inputs lead to better models. Start augmenting today.
🛠️ How to Use
1. 🐍 Python (Hugging Face Datasets)
pip install datasets
from datasets import load_dataset
# Load Q-Augmented
dataset = load_dataset("sKT-Ai-Labs/Q-Augmented")
# Inspect structure
print(dataset['train'][0])
# Example: Batch processing for evaluation
for batch in dataset['train']:
q1 = batch['question_1']
q2 = batch['question_2']
# Compare embeddings or test generation consistency
# ...
2. 🎯 Recommended Use Cases
| Use Case | Description |
|---|---|
| Semantic Similarity | Train embedding models to recognize equivalent questions. |
| Paraphrase Detection | Fine-tune classifiers for duplicate question detection. |
| Generation Diversity | Use as prompts to measure output variance in LLMs. |
| Curriculum Learning | Start with simple pairs, progress to complex augmentations. |
⚖️ License & Attribution
This dataset is released under the Apache-2.0 License.
- Created by: SKT AI LABS
- Source: Synthetically augmented from high-quality seed data.
- Attribution: Please cite "sKT-Ai-Labs/Q-Augmented" when used in research.
Made with ❤️ by SKT AI LABS
Building the foundation for next-gen AI reasoning.
- Downloads last month
- 133