aiqbench analysis
This study demonstrates how post-training via supervised fine-tuning and reinforcement learning, combined with iterative test-time inference, enables an LLM to solve Olympiad-level mathematics without external tools. By releasing the specialized checkpoints, training data, and a new benchmark of 200 problems, the authors provide a transparent pipeline that significantly advances the state-of-the-art in formal proof generation.
An Open Recipe for IMO Gold: Training Nemotron for Olympiad Mathematics
arXiv:2609.10712v1 Announce Type: new 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.
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.
Artificial Intelligence (cs.AI)
arXiv:2609.10712 [cs.AI]
(or arXiv:2609.10712v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2609.10712 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 239 words
Related coverage
One Model Family, Two Gold-Level Results: Fine-Tuning Nemotron for IOI and IMO
Hugging Face · Oct 7, 2026
[AINews] Agents disprove Polysemanticity; greatly improve Interpretability & Alignment
Latent Space · Oct 16, 2026
Scaling Indic Pretraining: 2.8 Trillion Tokens and an Open Data Generation Pipeline
Hugging Face · Oct 16, 2026
Introducing Instella-MoE: A State-of-the-Art Fully Open Mixture-of-Experts Language Model
AMD · Jul 24, 2026