README.md

metadata

license: cc-by-4.0
language:
  - en
task_categories:
  - text-generation
tags:
  - reasoning
  - math
  - code
  - science
  - supervised-fine-tuning
  - synthetic
pretty_name: NVIDIA Open Reasoning Balanced 100k
size_categories:
  - 10K<n<100K

nvidia_open_reasoning_balanced_100k

A domain-balanced 100k reasoning SFT dataset built from three NVIDIA Open Reasoning datasets: 33,333 examples each for math, code, and science (99,999 total).

Each example is a single-turn conversation with a full reasoning trace:

conversations: [
  {"from": "human", "value": "<problem>"},
  {"from": "gpt",   "value": "<think>\n<reasoning trace>\n</think><final solution>"}
]

Columns

column	description
conversations	human / gpt turns (gpt = <think>...</think> reasoning + solution)
domain	math / code / science (33,333 each)
source	origin dataset and subset, e.g. nvidia/OpenMathReasoning|aops_c6_high_school_olympiads, nvidia/OpenCodeReasoning|codeforces
difficulty	difficulty label where the origin dataset provides one (None otherwise)

Source datasets

This dataset is a balanced sample of the following NVIDIA datasets. All three are released under CC BY 4.0, and this dataset inherits that license.

domain	dataset	rows sampled
math	nvidia/OpenMathReasoning	33,333
code	nvidia/OpenCodeReasoning	33,333
science	nvidia/OpenScienceReasoning-2	33,333

Citation

This repository is only a convenience re-packaging. Please cite the original NVIDIA datasets, not this repository:

@article{moshkov2025aimo2,
  title   = {AIMO-2 Winning Solution: Building State-of-the-Art Mathematical Reasoning Models with OpenMathReasoning dataset},
  author  = {Ivan Moshkov and Darragh Hanley and Ivan Sorokin and Shubham Toshniwal and Christof Henkel and Benedikt Schifferer and Wei Du and Igor Gitman},
  year    = {2025},
  journal = {arXiv preprint arXiv:2504.16891}
}

@article{ahmad2025opencodereasoning,
  title   = {OpenCodeReasoning: Advancing Data Distillation for Competitive Coding},
  author  = {Wasi Uddin Ahmad and Sean Narenthiran and Somshubra Majumdar and Aleksander Ficek and Siddhartha Jain and Jocelyn Huang and Vahid Noroozi and Boris Ginsburg},
  year    = {2025},
  eprint  = {2504.01943},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CL},
  url     = {https://arxiv.org/abs/2504.01943}
}

For the science subset, cite the dataset page directly:

@misc{openscience_reasoning_2,
  title  = {OpenScienceReasoning-2},
  author = {NVIDIA},
  year   = {2025},
  howpublished = {\url{https://huggingface.co/datasets/nvidia/OpenScienceReasoning-2}}
}