An Open Recipe for IMO Gold: Training Nemotron for Olympiad Mathematics

Hugging Face Daily Papers Papers

Summary

The paper presents an open-method using post-trained Nemotron 3 Ultra checkpoints to achieve gold-medal performance on IMO 2026 through iterative verification and refinement in natural language without external tools.

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.
Original Article
View Cached Full Text

Cached at: 09/11/26, 06:16 AM

Paper page - An Open Recipe for IMO Gold: Training Nemotron for Olympiad Mathematics

Source: https://huggingface.co/papers/2609.10712

Abstract

A natural-language proof-generation pipeline using post-trained Nemotron 3 Ultra checkpoints achieves gold-medal performance on IMO 2026 through iterative verification and refinement without external tools.

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 usingsupervised fine-tuningandreinforcement learning, and evaluate checkpoint choice,verification, andrefinement. 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 aniterative searchthat 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, andNemotron-IMO-Bench, a new benchmark of 200 novel olympiad-level problems.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2609\.10712

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2609.10712 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2609.10712 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2609.10712 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Maxproof

Hacker News Top

MaxProof introduces a test-time scaling framework that combines proof generation, verification, and repair using generative-verifier RL, enabling the M3 model to exceed human gold-medal thresholds on IMO 2025 and USAMO 2026.