A high-quality dataset of mathematical reasoning preference pairs for Direct Preference Optimization (DPO) and Learning by Teaching (LBT) training. ✨ This dataset includes precomputed reference model log probabilities for efficient DPO training!

📊 Dataset Statistics

Total Examples: 44,565 Train Split: 37,895 Test Split: 6,670 Chapters: 30

	📚 Chapter Distribution

Chapter Count

3dgeometry 1947… See the full description on the dataset page: https://huggingface.co/datasets/AshwinKM2005/NCERTK12Math-LBT_DPO_Pairs.