Math-Glot-Cleaned is a filtered and structured version of the NVIDIA AceReason-1.1-SFT dataset, focusing specifically on high-quality math and code-related question-answer pairs. This version is intended for use in training and evaluating reasoning-focused large language models, particularly for math problem solving and structured generation tasks.
Source: Derived from NVIDIA’s AceReason-1.1-SFT
Size: 22,232 examples
Format:… See the full description on the dataset page:
https://huggingface.co/datasets/prithivMLmods/Math-Glot-Cleaned.
