Dataset Viewer
Auto-converted to Parquet Duplicate
expected_answer
stringlengths
0
1.62k
problem_type
stringclasses
3 values
problem_source
stringclasses
5 values
generation_model
stringclasses
2 values
pass_rate_72b_tir
stringclasses
33 values
problem
stringlengths
15
1.67k
generated_solution
stringlengths
1.02k
76k
inference_mode
stringclasses
1 value
used_in_kaggle
bool
2 classes
\(\frac{C_{n_1}^{a_1} \cdot C_{n_2}^{a_2} \cdot \ldots \cdot C_{n_C}^{a_C}}{C_N^A}\)
has_answer_extracted
aops_c6_high_school_olympiads
DeepSeek-R1
0.65625
Given a group of \( N \) balls consisting of \( C \) colors, where the number of balls in each color is represented as \( n_1, n_2, \ldots, n_C \) (with \( n_1 + n_2 + \ldots + n_C = N \)), what is the probability that when \( A \) balls are randomly picked (where \( A \leq N \)), the picked balls consist of \( a_1, a_...
<think> Okay, so I need to find the probability that when I pick A balls out of N, where there are C different colors, the number of each color I pick is exactly a1, a2, ..., aC. Hmm, let's think about how to approach this. First, probability problems often involve combinations. The general formula for probability is ...
cot
true
\frac{n(n-1)}{2}
no_answer_extracted
aops_c4_high_school_math
DeepSeek-R1
n/a
How many lines can be drawn that are equidistant from \( n \) given points in the plane?
<think> Okay, so the problem is asking how many lines can be drawn that are equidistant from n given points in the plane. Hmm, let me try to wrap my head around this. First, I need to recall what it means for a line to be equidistant from points. I think when we say a line is equidistant from two points, it's the per...
cot
true
\( f(x) = f(1)x \)
has_answer_extracted
aops_c6_high_school_olympiads
DeepSeek-R1
0.5625
Find all functions \( f: \mathbb{R} \to \mathbb{R} \) such that \[ (f(x^2) + f(xy) + f(y^2))(x - y) = f(x^3) - f(y^3) \] for all \( x, y \in \mathbb{R} \).
<think> Okay, let's try to solve this functional equation problem. The equation given is: \[ (f(x^2) + f(xy) + f(y^2))(x - y) = f(x^3) - f(y^3) \] We need to find all functions \( f: \mathbb{R} \to \mathbb{R} \) that satisfy this for all real numbers \( x \) and \( y \). Hmm, functional equations can be tricky, but m...
cot
true
2
has_answer_extracted
aops_c6_high_school_olympiads
DeepSeek-R1
0.625
Find the sum of the roots of the equation \((x+6)^{1/3} + (x-2)^{1/2} = 2\).
<think> Okay, let's see. I need to find the sum of the roots of the equation \((x+6)^{1/3} + (x-2)^{1/2} = 2\). Hmm, roots are the solutions for x that satisfy this equation. The sum of the roots would be all the solutions added together. But wait, how many roots does this equation have? Let me think. First, let's ana...
cot
true
32
has_answer_extracted
aops_c6_high_school_olympiads
QwQ-32B
0.0
Determine how many 1000 digit numbers \( A \) have the property that when any digit of \( A \), aside from the first, is deleted to form a 999 digit number \( B \), then \( B \) divides \( A \).
<think> Okay, so I need to figure out how many 109-digit numbers A have the property that if you delete any digit except the first one to form a 999-digit number B, then B divides A. Hmm, let's try to break this down step by step. First, let's make sure I understand the problem correctly. We have a 1000-digit number A...
cot
false
\(\frac{2}{\pi} + \frac{32}{9\pi^2}\)
has_answer_extracted
aops_c7_college_math
DeepSeek-R1
0.4375
Calculate the integral \[ \int^{\frac{3\pi}{2}}_{\frac{\pi}{2}} \left|\left(\frac{2}{x^3}+\frac{1}{x}\right)\sin x\right|dx \]
<think> Okay, let me try to solve this integral. So the problem is to compute the integral from π/2 to 3π/2 of the absolute value of (2/x³ + 1/x) times sin x dx. Hmm, that looks a bit intimidating at first glance, but let's break it down step by step. First, I remember that when dealing with integrals involving absolu...
cot
true
1
converted_proof
aops_c6_high_school_olympiads
QwQ-32B
n/a
In $\triangle ABC$ with incenter $I$, points $P$ and $Q$ are on side $BC$ such that $AP$ and $AQ$ are isogonal conjugates with respect to $\angle BAC$. The tangent line at $B$ of $\odot (ABP)$ and the tangent line at $C$ of $\odot (ACQ)$ intersect at $D$. Let $E = AP \cap IB$ and $F = AQ \cap IC$. If $DI$ intersects $E...
<think> Alright, let me try to tackle this geometry problem step by step. It's a bit complex, but I'll break it down as much as possible. First, let me restate the problem to make sure I understand it correctly. We have triangle ABC with incenter I. Points P and Q are on side BC such that AP and AQ are isogonal conjug...
cot
false
\(\frac{(3-\sqrt{3})(2-\sqrt{3})^{2016}+(3+\sqrt{3})(2+\sqrt{3})^{2016}}{6}\)
has_answer_extracted
aops_c6_high_school_olympiads
DeepSeek-R1
0.0
Let \( x_0 = 1 \), and \( x_{n+1} = 2x_n + \sqrt{3x_n^2 - 2} \). Determine \( x_{2016} \).
"<think>\nOkay, let's see. The problem is about a recursive sequence where x₀ is 1, and each subse(...TRUNCATED)
cot
true
$\sqrt[2012]{2013!} > \sqrt[2013]{2012!}$
has_answer_extracted
aops_c6_high_school_olympiads
DeepSeek-R1
0.0625
Which is greater, $\sqrt[2012]{2013!}$ or $\sqrt[2013]{2012!}$?
"<think>\nOkay, so I need to figure out whether the 2012th root of 2013 factorial is greater than th(...TRUNCATED)
cot
true
20
has_answer_extracted
aops_c6_high_school_olympiads
DeepSeek-R1
0.9375
On average, how long will you have to flip a coin to get the sequence $HTHT$?
"<think>\nOkay, so I need to figure out the expected number of coin flips needed to get the sequence(...TRUNCATED)
cot
true
End of preview. Expand in Data Studio

nvidia/OpenMathReasoning(複製)

データの内容

AoPS フォーラムの数学の問題に、DeepSeek-R1 や QwQ-32B が生成した解答を付けた、大規模な数学推論のデータ。1 行が 1 件の解答で、問題文は problem、生成された解答は generated_solution、正解は expected_answer に入る。解答は長い chain-of-thought、ツール統合推論(TIR)、多数の候補から選ぶ GenSelect の 3 種類で、問題の整形には Qwen2.5-32B-Instruct を使った。

データの概要

項目 値
行数 5,678,317(1 構成・4 分割の合計)
列数 9
容量 49.53 GB(ファイル 234 本)
ファイルの形式 parquet 198
言語 en
公開 public

構成と分割

構成(config) 分割(split)ごとの行数 列数
default cot 3,201,061、tir 1,718,466、genselect 565,620、additional_problems 193,170 9

列(構成 default・分割 cot)

列 型 先頭の行の値
expected_answer string \(\frac{C_{n_1}^{a_1} \cdot C_{n_2}^{a_2} \cdot …
problem_type string has_answer_extracted
problem_source string aops_c6_high_school_olympiads
generation_model string DeepSeek-R1
pass_rate_72b_tir string 0.65625
problem string Given a group of \( N \) balls consisting of \( …
generated_solution string <think> Okay, so I need to find the probability …
inference_mode string cot
used_in_kaggle bool True

行数と列は HF のデータセットビューアの集計(元のデータセットの現在の版で数えたもの)。全データセットの一覧は HYUKI5/dataset-catalog にある。

複製について

nvidia/OpenMathReasoning をリビジョン d3d08664755704f422af97d43a7ff0ded4bd95df の時点でそのまま複製したもの。中身は変えていない。判断モデル Laya(convaiinnovations/laya)の微調整に使う候補として、調査メモ(Laya 学習用データセット調査)に載せたデータを集めている。

出典

論文

アノテーションの方法

(元のカードに該当の見出しが無いので、ラベルの付け方に触れた段落を抜き出した)

  • problem: AoPS フォーラム から抽出し、Qwen2.5-32B-Instruct で整形した問題文。
  • generated_solution: DeepSeek-R1 または QwQ-32B を使って合成により生成された解答。
  • generation_model: 生成に使用したモデル。DeepSeek-R1 または QwQ-32B。
  • problem_type: "has_answer_extracted"、"no_answer_extracted"、"converted_proof" のいずれか。答えを抽出できたか、証明問題を答え付きの問題に変換したものかによって決まる。
  • expected_answer: "problem_type" が "has_answer_extracted" の場合は抽出された答え。それ以外の場合は、この問題について生成されたすべての解答の多数決による答え。
  • problem_source: 対応する AoPS フォーラム(例: "aops_c6_high_school_olympiads")または "MATH_training_set" を示す。後者は、MATH からの生成物も少数含むためである。
  • inference_mode: 推論モード。"cot"、"tir"、"genselect" のいずれか。
  • pass_rate_72b_tir: TIR モードで実行した Qwen2.5-Math-72B-Instruct の、32 回の生成のうちの合格率。この属性は "problem_type" が "has_answer_extracted" の場合にのみ利用でき、それ以外の場合は "n/a" となる。
  • used_in_kaggle: そのインスタンスが AIMO-2 Kaggle competition の優勝モデルの学習に使われたかどうか。OpenMath-Nemotron-14B-Kaggle モデルの学習には、2.2M 件の CoT 解答と 15K 件の TIR 解答を使用した。なお、OpenMath-Nemotron モデルの学習には、Kaggle で使用した TIR のサブセットを除くすべての CoT、TIR、GenSelect のデータを使用した。

微調整に使うときの注意

  • 調査メモの台帳での区分: A(許諾型のライセンスだけで、疑義の注記が無い)。区分は記載の文言で機械的に付けたもので、法的な判定ではない。学習に入れる前に元のカードとライセンスの原文で確かめる。
  • Laya の入力は 2,048 token まで(ModernBERT-large / mmBERT)。長い本文は先頭から黙って切られるので、切れる行は事前に除くか、要約・切り出しをしてから入れる。
  • 分割は記事・文書単位で行い、同じ本文から作った別の問いを学習側と較正側に分けない。温度と閾値は学習に使っていない分割で別々に当てる。

調査メモの記述

メモ 節 ライセンス(記載) 確認 注記
nvidia_nemotron.md 付録. 取得した nvidia 組織のデータセット全件の簡易一覧 (L367) cc-by-4.0 確認済 確認印なし(各ファイルの凡例により確認済扱い) / 原文注記: × 数学・コード・科学・計測

元のカード(日本語訳)

以下は元のデータセットのカードの本文を日本語に訳したもの(機械翻訳)。訳に疑問があるときは、末尾の原文を確かめる。


OpenMathReasoning

OpenMathReasoning は、大規模言語モデル(LLM)を学習させるための大規模な数学推論データセットである。 このデータセットには次が含まれる。

  • AoPS フォーラム から得た 306K 件の一意な数学の問題(次を含む):
    • 3.2M 件の長い chain-of-thought(CoT)解答
    • 1.7M 件の長いツール統合推論(TIR)解答
    • 566K 件の、多数の候補から最も有望な解答を選ぶサンプル(GenSelect)
  • AoPS フォーラムから得た追加の 193K 問(問題のみで解答はなし)

問題の前処理に Qwen2.5-32B-Instruct を、解答の生成に DeepSeek-R1 と QwQ-32B を使用した。

このデータセットは、AIMO-2 Kaggle competition での我々の優勝提出の基盤となった。

詳細は 論文 を参照のこと。

注記: 当初、データセットには 540K の一意な問題が含まれると報告していたが、この数字はパイプラインの開始時点の問題数を表していた。公開されたデータセットに含まれる実際の CoT と TIR の解答は 306K 問に対応する。OpenMath-Nemotron モデルはこの縮小したサブセットで学習されているため、公開されたすべての結果は現在のリリースで再現可能である。誤っていたのは当初の問題数のみである。

問題数のこの食い違いは、2 つの要因で説明される。

  • フィルタリングの過程で、形式の制約(yes/no 問題、多肢選択問題など)、ベンチマークの汚染除去、既存の LLM が特定の問題に対して有効な解答を生成できないこと、により多くの問題を除いた。
  • パイプラインのバグにより、証明問題 137K 問を失った。復旧したこの追加データを学習に含めたところ、SFT の性能が低下した。復旧した問題を取り込む方法を現在検証しており、明確な性能向上が確認できた場合にのみ、その解答を公開する。

注記: このデータの初期版は、Llama-Nemotron-Post-Training-Dataset で別途公開された。

データセットのフィールド

OpenMathReasoning データセットには、次のフィールドが含まれる:

  • problem: AoPS フォーラム から抽出し、Qwen2.5-32B-Instruct で整形した問題文。
  • generated_solution: DeepSeek-R1 または QwQ-32B を使って合成により生成された解答。
  • generation_model: 生成に使用したモデル。DeepSeek-R1 または QwQ-32B。
  • problem_type: "has_answer_extracted"、"no_answer_extracted"、"converted_proof" のいずれか。答えを抽出できたか、証明問題を答え付きの問題に変換したものかによって決まる。
  • expected_answer: "problem_type" が "has_answer_extracted" の場合は抽出された答え。それ以外の場合は、この問題について生成されたすべての解答の多数決による答え。
  • problem_source: 対応する AoPS フォーラム(例: "aops_c6_high_school_olympiads")または "MATH_training_set" を示す。後者は、MATH からの生成物も少数含むためである。
  • inference_mode: 推論モード。"cot"、"tir"、"genselect" のいずれか。
  • pass_rate_72b_tir: TIR モードで実行した Qwen2.5-Math-72B-Instruct の、32 回の生成のうちの合格率。この属性は "problem_type" が "has_answer_extracted" の場合にのみ利用でき、それ以外の場合は "n/a" となる。
  • used_in_kaggle: そのインスタンスが AIMO-2 Kaggle competition の優勝モデルの学習に使われたかどうか。OpenMath-Nemotron-14B-Kaggle モデルの学習には、2.2M 件の CoT 解答と 15K 件の TIR 解答を使用した。なお、OpenMath-Nemotron モデルの学習には、Kaggle で使用した TIR のサブセットを除くすべての CoT、TIR、GenSelect のデータを使用した。

OpenMath-Nemotron モデル

このデータセットの品質を示すため、このデータで学習させた一連の OpenMath-Nemotron モデルを公開する。

Evaluation Results

これらのモデルは、代表的な数学ベンチマークで最先端の結果を達成している。指標は pass@1 (maj@64) で示す。ここで pass@1 は 64 回の生成の平均精度、maj@64 は多数決の結果である。評価の設定の詳細は 論文 を参照のこと。

Model AIME24 AIME25 HMMT-24-25 HLE-Math
DeepSeek-R1-Distill-Qwen-1.5B 26.8 (60.0) 21.4 (36.7) 14.2 (26.5) 2.9 (5.0)
OpenMath-Nemotron-1.5B CoT 61.6 (80.0) 49.5 (66.7) 39.9 (53.6) 5.4 (5.4)
OpenMath-Nemotron-1.5B TIR 52.0 (83.3) 39.7 (70.0) 37.2 (60.7) 2.5 (6.2)
+ 自己 GenSelect 83.3 70.0 62.2 7.9
+ 32B GenSelect 83.3 70.0 62.8 8.3
DeepSeek-R1-Distill-Qwen-7B 54.4 (80.0) 38.6 (53.3) 30.6 (42.9) 3.3 (5.2)
OpenMath-Nemotron-7B CoT 74.8 (80.0) 61.2 (76.7) 49.7 (57.7) 6.6 (6.6)
OpenMath-Nemotron-7B TIR 72.9 (83.3) 57.5 (76.7) 54.6 (66.3) 7.8 (10.8)
+ 自己 GenSelect 86.7 76.7 68.4 11.5
+ 32B GenSelect 86.7 76.7 69.9 11.9
DeepSeek-R1-Distill-Qwen-14B 65.8 (80.0) 48.4 (60.0) 40.1 (52.0) 4.2 (4.8)
OpenMath-Nemotron-14B-MIX (kaggle) 73.7 (86.7) 57.9 (73.3) 50.5 (64.8) 5.7 (6.5)
OpenMath-Nemotron-14B CoT 76.3 (83.3) 63.0 (76.7) 52.1 (60.7) 7.5 (7.6)
OpenMath-Nemotron-14B TIR 76.3 (86.7) 61.3 (76.7) 58.6 (70.9) 9.5 (11.5)
+ 自己 GenSelect 86.7 76.7 72.4 14.1
+ 32B GenSelect 90.0 76.7 71.9 13.7
QwQ-32B 78.1 (86.7) 66.5 (76.7) 55.9 (63.3) 9.0 (9.5)
DeepSeek-R1-Distill-Qwen-32B 66.9 (83.3) 51.8 (73.3) 39.9 (51.0) 4.8 (6.0)
OpenMath-Nemotron-32B CoT 76.5 (86.7) 62.5 (73.3) 53.0 (59.2) 8.3 (8.3)
OpenMath-Nemotron-32B TIR 78.4 (93.3) 64.2 (76.7) 59.7 (70.9) 9.2 (12.5)
+ 自己 GenSelect 93.3 80.0 73.5 15.7
DeepSeek-R1 79.1 (86.7) 64.3 (73.3) 53.0 (59.2) 10.5 (11.4)

結果の再現

データとモデルの作成に使ったパイプラインは、完全にオープンソースで公開されている。

データ生成を含め、結果を完全に再現するための すべての手順 を提供する。

引用

本研究が役立った場合は、引用を検討すること。

@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}
}

データセットの所有者:

NVIDIA Corporation

公開日:

04/23/2025

データのバージョン

1.0 (04/23/2025)

ライセンス/利用規約:

cc-by-4.0

想定される用途:

このデータセットは、コミュニティがモデルを改善し続けるために使われることを想定している。このデータは学習と評価に自由に使ってよい。

倫理的配慮:

NVIDIA は、信頼できる AI は共有の責任であると考えており、幅広い AI アプリケーションの開発を可能にするための方針と慣行を定めている。利用規約に従ってダウンロード・利用する場合、開発者は社内のモデルチームと協力し、関連する業界と用途の要件を満たしていることを確認し、想定外の製品の誤用に対処すること。

セキュリティの脆弱性や NVIDIA AI に関する懸念は、こちら から報告すること。

原文(英語)

OpenMathReasoning

OpenMathReasoning is a large-scale math reasoning dataset for training large language models (LLMs). This dataset contains

  • 306K unique mathematical problems sourced from AoPS forums with:
    • 3.2M long chain-of-thought (CoT) solutions
    • 1.7M long tool-integrated reasoning (TIR) solutions
    • 566K samples that select the most promising solution out of many candidates (GenSelect)
  • Additional 193K problems sourced from AoPS forums (problems only, no solutions)

We used Qwen2.5-32B-Instruct to preprocess problems, and DeepSeek-R1 and QwQ-32B to generate solutions.

This dataset was a foundation of our winning submission to the AIMO-2 Kaggle competition.

See our paper to learn more details!

NOTE: We initially reported 540K unique problems in our dataset, but this figure represented the question count at the pipeline's beginning. But the actual CoT and TIR solutions in the released dataset correspond to 306K problems. Since our OpenMath-Nemotron models were trained on this reduced subset, all published results remain reproducible with the current release—only our initial problem count was overstated.

Two factors explain this discrepancy in question numbers.

  • Our filtering process removed many questions due to format restrictions (yes/no questions, multiple choice questions, etc.), benchmark decontamination, and the inability of existing LLMs to generate valid solutions for certain problems.
  • A pipeline bug caused us to lose 137K proof-based questions. When we recovered and included this additional data in training, the SFT performance regressed. We are currently testing different approaches for incorporating these recovered questions and will release their solutions only if we identify clear performance improvements.

NOTE: An early version of this data was released separately in Llama-Nemotron-Post-Training-Dataset.

Dataset fields

OpenMathReasoning dataset contains the following fields:

  • problem: Problem statement extracted from AoPS forums and refined with Qwen2.5-32B-Instruct
  • generated_solution: Synthetically generated solution using either DeepSeek-R1 or QwQ-32B
  • generation_model: DeepSeek-R1 or QwQ-32B
  • problem_type: Can be one of "has_answer_extracted", "no_answer_extracted" and "converted_proof" dependening on whether we were able to extract the answer or if this is a proof question converted to answer question.
  • expected_answer: Extracted answer if "problem_type" is "has_answer_extracted". Otherwise this is the majority-voting answer across all generated solutions for this problem.
  • problem_source: States the corresponding AoPS forum (e.g. "aops_c6_high_school_olympiads") or "MATH_training_set" as we also include a small set of generations from MATH.
  • inference_mode: "cot", "tir" or "genselect"
  • pass_rate_72b_tir: Pass rate out of 32 generations for Qwen2.5-Math-72B-Instruct run in TIR mode. This attribute is only available when "problem_type" is "has_answer_extracted" and is set to "n/a" for other cases.
  • used_in_kaggle: Whether the instance was used in training the winning model for AIMO-2 Kaggle competition or not. We had used 2.2M CoT and 15K TIR solutions for training the OpenMath-Nemotron-14B-Kaggle model. Note that for training the OpenMath-Nemotron models, we used all the CoT, TIR, and GenSelect data, except for the TIR subset used in Kaggle.

OpenMath-Nemotron models

To demonstrate the quality of this dataset, we release a series of OpenMath-Nemotron models trained on this data.

Evaluation Results

The models achieve state-of-the-art results on popular mathematical benchmarks. We present metrics as pass@1 (maj@64) where pass@1 is an average accuracy across 64 generations and maj@64 is the result of majority voting. Please see our paper for more details on the evaluation setup.

Model AIME24 AIME25 HMMT-24-25 HLE-Math
DeepSeek-R1-Distill-Qwen-1.5B 26.8 (60.0) 21.4 (36.7) 14.2 (26.5) 2.9 (5.0)
OpenMath-Nemotron-1.5B CoT 61.6 (80.0) 49.5 (66.7) 39.9 (53.6) 5.4 (5.4)
OpenMath-Nemotron-1.5B TIR 52.0 (83.3) 39.7 (70.0) 37.2 (60.7) 2.5 (6.2)
+ Self GenSelect 83.3 70.0 62.2 7.9
+ 32B GenSelect 83.3 70.0 62.8 8.3
DeepSeek-R1-Distill-Qwen-7B 54.4 (80.0) 38.6 (53.3) 30.6 (42.9) 3.3 (5.2)
OpenMath-Nemotron-7B CoT 74.8 (80.0) 61.2 (76.7) 49.7 (57.7) 6.6 (6.6)
OpenMath-Nemotron-7B TIR 72.9 (83.3) 57.5 (76.7) 54.6 (66.3) 7.8 (10.8)
+ Self GenSelect 86.7 76.7 68.4 11.5
+ 32B GenSelect 86.7 76.7 69.9 11.9
DeepSeek-R1-Distill-Qwen-14B 65.8 (80.0) 48.4 (60.0) 40.1 (52.0) 4.2 (4.8)
OpenMath-Nemotron-14B-MIX (kaggle) 73.7 (86.7) 57.9 (73.3) 50.5 (64.8) 5.7 (6.5)
OpenMath-Nemotron-14B CoT 76.3 (83.3) 63.0 (76.7) 52.1 (60.7) 7.5 (7.6)
OpenMath-Nemotron-14B TIR 76.3 (86.7) 61.3 (76.7) 58.6 (70.9) 9.5 (11.5)
+ Self GenSelect 86.7 76.7 72.4 14.1
+ 32B GenSelect 90.0 76.7 71.9 13.7
QwQ-32B 78.1 (86.7) 66.5 (76.7) 55.9 (63.3) 9.0 (9.5)
DeepSeek-R1-Distill-Qwen-32B 66.9 (83.3) 51.8 (73.3) 39.9 (51.0) 4.8 (6.0)
OpenMath-Nemotron-32B CoT 76.5 (86.7) 62.5 (73.3) 53.0 (59.2) 8.3 (8.3)
OpenMath-Nemotron-32B TIR 78.4 (93.3) 64.2 (76.7) 59.7 (70.9) 9.2 (12.5)
+ Self GenSelect 93.3 80.0 73.5 15.7
DeepSeek-R1 79.1 (86.7) 64.3 (73.3) 53.0 (59.2) 10.5 (11.4)

Reproducing our results

The pipeline we used to produce the data and models is fully open-sourced!

We provide all instructions to fully reproduce our results, including data generation.

Citation

If you find our work useful, please consider citing us!

@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}
}

Dataset Owner(s):

NVIDIA Corporation

Release Date:

04/23/2025

Data Version

1.0 (04/23/2025)

License/Terms of Use:

cc-by-4.0

Intended Usage:

This dataset is intended to be used by the community to continue to improve models. The data may be freely used to train and evaluate.

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please report security vulnerabilities or NVIDIA AI Concerns here.

Downloads last month
50

Paper for HYUKI5/nvidia__OpenMathReasoning