| title | An Open Recipe for IMO Gold: Training Nemotron for Olympiad Mathematics |
|---|---|
| authors | Ivan Moshkov, Stephen Ge, George Armstrong, Wei Du, Sadegh Mahdavi, Igor Gitman |
| date | 2026-09-09 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2609.10712 |
| affiliation | NVIDIA |
| upvotes | 21 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf |
| abstract | We study how model post-training and test-time inference design affect natural-language proof generation for hard olympiad mathematics. Starting from Nemotron 3 Ultra, we train two specialist checkpoints using supervised fine-tuning and reinforcement learning, and evaluate checkpoint choice, verification, and refinement. Based on these findings, we present an open-model test-time-compute pipeline. The system operates entirely in natural language, with no formal prover, external tools, or internet access. Three Nemotron 3 Ultra checkpoints - the general-availability model and two post-trained specialists - power an iterative search that generates, verifies, and refines candidate proofs; a separate high-compute stage then selects each final submission. The system scored 30 out of 42 points at IMO 2026, reaching the gold-medal threshold. We release the two post-trained checkpoints as well as the training data, the training and inference code, the submitted solutions, and Nemotron-IMO-Bench, a new benchmark of 200 novel olympiad-level problems. |
cite
@misc{moshkov2026open,
title={An Open Recipe for IMO Gold: Training Nemotron for Olympiad Mathematics},
author={Ivan Moshkov and Stephen Ge and George Armstrong and Wei Du and Sadegh Mahdavi and Igor Gitman},
year={2026},
eprint={2609.10712},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2609.10712}
}