一个模范家庭，两个黄金级的结果：国际奥运会和海事组织的微调 Nemotron

一个模范家庭，两个黄金级的结果：国际奥运会和海事组织的微调 Nemotron

我们最近的结果表明，尼莫特龙是建设世界级专家模型的坚实、适应性强的基础。 从Nemotron 3开始，我们的团队使用监督的微调(SFT)，强化学习(RL)，以及反馈驱动的推论，在海事组织2026年和IOI 2026年创建了达到金介级的系统。

IOI的结果来自同时期的现场直播，前景预测，互联网接入，以及作为人类参赛者的提交限制。 它是非官方的，无人监督的基准测试，没有被列入IOI的官方排名。 海事组织系统提交的证明由海事组织正式分级人员分级。

可重复使用的专业化方法

"方便到微调"应该比使检查站可以列车更有意义。 这应当意味着，一个有能力的基础模型可以适应一个要求很高的领域，有明确、可重复使用的办法。

在整个两个项目中，这种配方有四个部分：

- 从强大的尼莫特龙基地模型开始

- Curate域特有问题和高质量的推理痕迹。

- 应用标准的培训后方法，如SFT，并在有用的地方采用RL。

- 用一个推论循环对专家模型进行比对，从而生成、评价和改进候选人的答案。

培训和推论是实质性的，但基本做法是熟悉的，可以复制。 我们不需要为每一项挑战建立一个新的基础模式。 我们专攻尼莫特隆的任务

从一般编码能力到IOI黄金

对于竞争性编程，我们研究了22,000个问题，并产生了合成推理痕迹来培养两名专家。 内莫特龙-3-纳诺-CC总参数300亿，活性参数30亿，同时获得SFT和RL. 内莫特龙-3-Ultra-CC总参数5500亿，活性参数550亿，均获得SFT.

2025年IOI的进展使得专业化的价值易于看到。 纳诺从训练后130分提高到SFT后280分，RL后291分。 随着GenCorrect，我们的迭代生成-评价-refine策略，它达到了468分，并超越了438.3的黄金门槛。 超CC达到502分，采用相同的测试时间策略。

这些实验还表明，适应不必在每一个尺度上都看一样。 SFT产生了纳诺的大部分收益，RL增加了一个较小但一致的改进。 对于更强大的Ultra模型来说，一个SFT时代足以超越IOI,CPC，和LiveCode Bench Pro全程训练的Nano模型。 这一调查结果指导了用于IOI 2026的针对竞争的超CC系统，在600分中达到535.4分。

教尼莫特龙证明、检查和修改

海事组织项目在奥林匹亚数学方面采用了同样的想法。 从Nemotron 3 Ultra开始，我们训练了一名专家与SFT，另一名与RL。

SFT程序包括414,890个质量过滤的例子，涉及15,818个独特的证据问题。 它所做的不只是教最后的答案。 数据涵盖证明的生成、完善、核实和meta核查，因此该模型学会了构建论据、找出差距、回应批评和判断证明是否完整。 RL模型在模型能力前沿附近选取的9,597个证明问题接受了培训。

在开发试验中，经过培训后的两个检查站都超过了一般可用模式。 SFT检查站在第一轮搜查中最强，而RL检查站则取得了最佳的总体单一检查点结果。 他们的长处是相辅相成的，因此最后的系统除了一般模式外，还使用了两位专家。

对于海事组织的每一个问题，模型都产生候选证明，评分，提出批评，并改进最有希望的尝试。 一个单独的高计算阶段选定了最后划界案。 整个系统使用自然语言，没有正式的证明、外部工具或互联网。 它在42分中得分30分，包括对6个问题中的4个问题给予完全的肯定，并超过了官方的金庸门槛。

微调和测试时间计算工作

我们先前的IOI 2025 Hugging Face文章显示，测试时间计算可以如何将开放量级模型推向金级性能。 新的结果又增加了一个重要的作品：更好的专业化使推论系统有更好的候选人，更好的批评，更好的完善。

在IOI中，GenCorrect将微调的收益转化为比多个反馈回合更大的改进。 在海事组织，使用补充的SFT和RL检查站比从一个检查站抽取更多样本更有价值。 在这两种情况下，最好的结果都是将有能力的专家与能够搜索、核实和改进的系统结合起来。

这种区别很重要。 奖章并非仅由微调生产，也并非仅由畜力采样生产。 他们来自共同设计模型，数据，和推论循环。

Hugging Face上的开放模型、数据和食谱

我们希望这些结果在竞争之外有用。 海事组织的Nemotron Labs 2026年收集带来，同时收集SFT和RL检查站，两者都是培训数据集，Nemotron-IMO-Bench，一个新的基准测试，涉及200个奥林匹克级问题。 海事组织的文件描述了培训办法和生成-核实-修复系统，而NeMo-Skills储存库包括海事组织的推论管道、提示、提交的证据以及可复制的快速启动。 对于竞争性编程，可在Hugging Face上获得Nemotron-3-Ultra-CC模型，IOI文件提供了培训配方和GenCorrect方法。 NeMo-Skills也提供IOI评价和推论管道。

海事组织和IOI共同为一个简单的想法提供了异常苛刻的证据：Nemotron可以被微调成世界级域专家，然后由透明的推论工作流程组成，以解决人类竞争前沿的问题。

我们很高兴看到Hugging Face社区接下来会建设什么。

本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。

One Model Family, Two Gold-Level Results: Fine-Tuning Nemotron for IOI and IMO

Our recent results show that Nemotron is a strong, adaptable foundation for building world-class specialist models. Starting from Nemotron 3, our teams used supervised fine-tuning (SFT), reinforcement learning (RL), and feedback-driven inference to create systems that reached gold-medal level at both IMO 2026 and IOI 2026.

The IOI result came from a live, prospective run under the same time, internet-access, and submission constraints as human contestants. It was an unofficial, unsupervised benchmark and was not included in the official IOI ranking. The IMO system’s submitted proofs were graded by official IMO graders.

A reusable specialization recipe

"Easy to fine-tune" should mean more than making a checkpoint trainable. It should mean that a capable foundation model can be adapted to a demanding domain with a clear, reusable recipe.

Across the two projects, that recipe had four parts:

- Start with a strong Nemotron base model.

- Curate domain-specific problems and high-quality reasoning traces.

- Apply standard post-training methods such as SFT and, where useful, RL.

- Pair the specialist model with an inference loop that generates, evaluates, and improves candidate answers.

The training and inference runs were substantial, but the underlying approach is familiar and reproducible. We did not need to build a new foundation model for every challenge. We specialized Nemotron for the task.

From general coding ability to IOI gold

For competitive programming, we curated 22,000 problems and generated synthetic reasoning traces to train two specialists. Nemotron-3-Nano-CC, with 30 billion total parameters and 3 billion active parameters, received both SFT and RL. Nemotron-3-Ultra-CC, with 550 billion total parameters and 55 billion active parameters, received SFT.

The progression on IOI 2025 makes the value of specialization easy to see. Nano improved from 130 points before post-training to 280 after SFT and 291 after RL. With GenCorrect, our iterative generate-evaluate-refine strategy, it reached 468 points and crossed the gold threshold of 438.3. Ultra-CC reached 502 points with the same test-time strategy.

These experiments also showed that adaptation does not have to look the same at every scale. SFT produced most of Nano's gain, with RL adding a smaller but consistent improvement. For the stronger Ultra model, one SFT epoch was enough to outperform the fully post-trained Nano model across IOI, ICPC, and LiveCodeBench Pro. That finding guided the competition-specific Ultra-CC system used for IOI 2026, which scored 535.4 out of 600.

Teaching Nemotron to prove, check, and revise

The IMO project applied the same idea to olympiad mathematics. Starting from Nemotron 3 Ultra, we trained one specialist with SFT and another with RL.

The SFT corpus contained 414,890 quality-filtered examples across 15,818 unique proof problems. It did more than teach final answers. The data covered proof generation, refinement, verification, and meta-verification, so the model learned to construct arguments, identify gaps, respond to critiques, and judge whether a proof was complete. The RL model was trained on 9,597 proof problems selected near the model's capability frontier.

Both post-trained checkpoints outperformed the general-availability model in the development experiments. The SFT checkpoint was strongest in the first search round, while the RL checkpoint achieved the best overall single-checkpoint result. Their strengths were complementary, so the final system used both specialists alongside the general model.

For each IMO problem, the models generated candidate proofs, scored them, produced critiques, and refined the most promising attempts. A separate high-compute stage selected the final submission. The entire system worked in natural language, with no formal prover, external tools, or internet access. It scored 30 out of 42 points, including full credit on four of the six problems, and exceeded the official gold-medal threshold.

Fine-tuning and test-time compute work together

Our earlier IOI 2025 Hugging Face post showed how test-time compute can push open-weight models to gold-level performance. The new results add an important piece: better specialization gives the inference system better candidates, better critics, and better refinements.

At IOI, GenCorrect turned the gains from fine-tuning into larger improvements over multiple feedback rounds. At IMO, using complementary SFT and RL checkpoints was more valuable than simply drawing more samples from one checkpoint. In both cases, the best outcome came from combining a capable specialist with a system that could search, verify, and improve.

This distinction matters. The medals were not produced by fine-tuning alone, and they were not produced by brute-force sampling alone. They came from co-designing the model, the data, and the inference loop.

Open models, data, and recipes on Hugging Face

We want these results to be useful beyond the competitions. The Nemotron Labs IMO 2026 collection brings together the SFT and RL checkpoints, both training datasets, and Nemotron-IMO-Bench, a new benchmark of 200 olympiad-level problems. The IMO paper describes the training approach and generate-verify-refine system, while the NeMo-Skills repository includes the IMO inference pipeline, prompts, submitted proofs, and a reproducible quickstart. For competitive programming, the Nemotron-3-Ultra-CC model is available on Hugging Face, and the IOI paper provides the training recipe and the GenCorrect methodology. The IOI evaluation and inference pipeline are also available in NeMo-Skills.

Together, IMO and IOI provide unusually demanding evidence for a simple idea: Nemotron can be fine-tuned into world-class domain specialists, then composed with transparent inference workflows to solve problems at the frontier of human competition.

We are excited to see what the Hugging Face community builds next.

Auto-fetched and translated by AI Atlas. All rights belong to the original author.