Nemotron-SFT-Science-v2

Dataset Description: Nemotron-Science-v2 is a science reasoning dataset with synthetic (synthetic MCQ, RQA) and non-synthetic vendor problems and LLM-generated solutions. It comprises three domains (Physics, Biology, and Chemistry), two question formats (multiple-choice questions [MCQ] and open questions [OpenQ]), and three generation setups: chain-of-thought (CoT) reasoning without tools, Python tool usage, and search tools usage with the Tavily API. The solutions were generated with open-source LLMS: GPT-OSS, Kimi-K2-Instruct, DeepSeek-V3.2, and DeepSeek-V4-Pro. There are four subsets of the data: Synthetic MCQ**: Synthetically generated problems with 4-10 multiple choice options. RQA**: Synthetic Chemistry problems (both MCQ and OpenQ). Vendor**: Non-synthetic, graduate/research-level open-ended STEM problems spanning Physics, Chemistry, and Biology. SO MCQ**: Multiple-choice questions derived from Stack Exchange covering physics, biology, and chemistry domains. This dataset is ready for commercial or non-commercial uses. Dataset Owner(s): NVIDIA Corporation Dataset Creation Date: Created on: May 13, 2026 Last Modified on: May 13, 2026 Version: Nemotron-SFT-Science-v2 This is a replacement to the previous dataset. Previous Version(s): https://huggingface.co/datasets/nvidia/Nemotron-Science-v1 License/Terms of Use: CC BY-SA 4.0 Intended Usage: This dataset is intended for LLM engineers and research teams developing and training large language models with a focus on improving scientific reasoning and problem-solving capabilities. It is suitable for supervised training and data augmentation in science-based model development pipelines. Dataset Characterization Data Collection Method** Hybrid: Synthetic, Automated Labeling Method** Synthetic Dataset Format Modality: Text Format: JSONL Structure: Text + Metadata Dataset Quantification | File | Rows | Size | |---|---|---| | Synthetic MCQ | 96,610 | 0.57 GB | | Vendor | 221,598 | 25 GB | | RQA | 413,519 | 8.9 GB | | SO|