File size: 14,628 Bytes
26da209
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
---
license: cc-by-nc-4.0
pretty_name: nvidia/AceMath-Instruct-Training-Data (mirror)
source_datasets:
- nvidia/AceMath-Instruct-Training-Data
tags:
- laya
- laya-finetune
- mirror
- domain-math
- domain-code
- use-noncommercial
- genre-instruction
- label-text
- annot-llm
- lang-en
language:
- en
configs:
- config_name: default
  data_files:
  - split: general_sft_stage1
    path: data/general_sft_stage1.parquet
  - split: general_sft_stage2
    path: data/general_sft_stage2.parquet
  - split: math_sft
    path: data/math_sft.parquet
dataset_info:
  features:
  - name: messages
    list:
    - name: role
      dtype: string
    - name: content
      dtype: string
  - name: answer
    dtype: string
  splits:
  - name: general_sft_stage1
  - name: general_sft_stage2
  - name: math_sft
---

# nvidia/AceMath-Instruct-Training-Data(複製)

## データの内容

AceMath-Instruct モデルの学習に使った、数学・コード・一般領域の指示チューニング用のデータ。1 行が 1 件の会話で、発話の並び `messages` と正解 `answer` を持ち、汎用の SFT 2 段階(general_sft_stage1・stage2)と数学の SFT(math_sft)の 3 つに分かれる。数学のプロンプトへの応答は Qwen2.5-Math-72B-Instruct、それ以外は GPT-4o-mini で生成した。

## データの概要

| 項目 | 値 |
|---|---|
| 行数 | 5,557,354(1 構成・3 分割の合計) |
| 列数 | 2 |
| 容量 | 5.86 GB(ファイル 6 本) |
| ファイルの形式 | parquet 3、png 1 |
| 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` | `general_sft_stage1` 2,261,687、`general_sft_stage2` 1,634,573、`math_sft` 1,661,094 | 2 |

### 列(構成 `default`・分割 `general_sft_stage1`)

| 列 | 型 | 先頭の行の値 |
|---|---|---|
| `messages` | `list<struct{role: string, content: string}>` | `[1 件]` |
| `answer` | `string` | `To determine the area of triangle \( DAB \) give…` |

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](https://huggingface.co/datasets/HYUKI5/dataset-catalog) にある。

## 複製について

[`nvidia/AceMath-Instruct-Training-Data`](https://huggingface.co/datasets/nvidia/AceMath-Instruct-Training-Data) をリビジョン `24e49e73a3ab3aa373e904516838c32a972ba75e` 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

- 元のデータセット: https://huggingface.co/datasets/nvidia/AceMath-Instruct-Training-Data
- 複製したリビジョン: `24e49e73a3ab3aa373e904516838c32a972ba75e`(2025-01-17 12:41:19+00:00)
- 元のライセンス表記: `cc-by-nc-4.0`

## 論文

- https://arxiv.org/abs/2412.15084

## 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)

AceMath-Instruct の学習データセットは <b>Built with Qwen</b> である。数学のプロンプトの出力は Qwen2.5-Math-72B-Instruct で生成した。それ以外の種類のプロンプトの出力は GPT-4o-mini で生成した。

**ライセンス**
AceMath-Instruct の学習データセットは非商用利用に限られ、OpenAI が生成したデータの [利用規約](https://openai.com/policies/row-terms-of-use/) に従う。このデータセットは、[Creative Commons Attribution: Non-Commercial 4.0 International](https://spdx.org/licenses/CC-BY-NC-4.0) のライセンスの下で提供する。

## 微調整に使うときの注意

- 調査メモの台帳での区分: C(NC・ND・研究専用・評価専用など、商用の利用を妨げうる記載がある)。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 (L387) | cc-by-nc-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。

---

[ウェブサイト](https://research.nvidia.com/labs/adlr/acemath/) | [論文](https://arxiv.org/abs/2412.15084)


## AceMath-Instruct 学習データのカード
AceMath-1.5B/7B/72B-Instruct モデルを学習させるためのデータセットをすべて公開する。これらのモデルは、多段階の教師あり微調整(SFT)の過程を経て、Qwen2.5-Math-Base モデルを基に作られている。微調整は汎用の SFT データ(general_sft_stage1.parquet と general_sft_stage2.parquet)から始まり、数学に特化した SFT データ(math_sft.parquet)が続く。私たちの実験では、数学に特化した SFT データだけで Qwen2.5-Math-Base モデルを微調整しても、競争力のある数学の推論性能が得られた。

AceMath-Instruct の学習データセットは <b>Built with Qwen</b> である。数学のプロンプトの出力は Qwen2.5-Math-72B-Instruct で生成した。それ以外の種類のプロンプトの出力は GPT-4o-mini で生成した。

データの統計は次のとおり:
- general_sft_stage1: 2,261,687 件のサンプル(コードと数学の SFT サンプルからなる)
- general_sft_stage2: 1,634,573 件のサンプル(コードと数学と一般領域の SFT サンプルからなる)
- math_sft: 1,661,094 件(数学の SFT サンプルのみ)


## ベンチマーク結果(AceMath-Instruct + AceMath-72B-RM)

<p align="center">
  <img src="./acemath-pic.png" alt="AceMath ベンチマーク結果" width="800">
</p>


上の表で、AceMath を主要な独自モデルと公開モデルの数学モデルと比較している。AceMath-7B-Instruct は、さまざまな数学の推論ベンチマークで、従来の最良の Qwen2.5-Math-7B-Instruct を大きく上回る(平均 pass@1: 67.2 対 62.9)。一方で、10 倍の規模の Qwen2.5-Math-72B-Instruct の性能に近づいている(67.2 対 68.2)。特筆すべきは、AceMath-72B-Instruct が、最先端の Qwen2.5-Math-72B-Instruct(71.8 対 68.2)、GPT-4o(67.4)、Claude 3.5 Sonnet(65.6)を大きな差で上回ったことである。また、報酬モデル AceMath-72B-RM が達成した rm@8 の正解率(8 つのうち最良)も報告する。これは、これらの推論ベンチマークで新しい記録を打ち立てている。ただし、スケールされた推論計算に頼る OpenAI の o1 モデルは含めていない。


## 使い方
```python
from datasets import load_dataset

data_files = {
    "general_sft_stage1": "data/general_sft_stage1.parquet",
    "general_sft_stage2": "data/general_sft_stage2.parquet",
    "math_sft": "data/math_sft.parquet",
}

# load the datasets
dataset = load_dataset(
    "nvidia/AceMath-Instruct-Training-Data",
    data_files=data_files,
    cache_dir="CACHE_DIR_TO_STORE_THE_DATASET"
)

# print example
print(dataset['math_sft'][0])

# example format
"""
{
    "messages": [
        {
            "role": "user",
            "content": "...",
        }
    ],
    "answer": "..."
}
"""
```


## すべてのリソース
### AceMath の指示調整モデル
- [AceMath-1.5B-Instruct](https://huggingface.co/nvidia/AceMath-1.5B-Instruct), [AceMath-7B-Instruct](https://huggingface.co/nvidia/AceMath-7B-Instruct), [AceMath-72B-Instruct](https://huggingface.co/nvidia/AceMath-72B-Instruct)

### AceMath の報酬モデル
- [AceMath-7B-RM](https://huggingface.co/nvidia/AceMath-7B-RM), [AceMath-72B-RM](https://huggingface.co/nvidia/AceMath-72B-RM)

### 評価と学習のデータ
- [AceMath-RewardBench](https://huggingface.co/datasets/nvidia/AceMath-RewardBench), [AceMath-Instruct Training Data](https://huggingface.co/datasets/nvidia/AceMath-Instruct-Training-Data), [AceMath-RM Training Data](https://huggingface.co/datasets/nvidia/AceMath-RM-Training-Data)

### 汎用の指示調整モデル
- [AceInstruct-1.5B](https://huggingface.co/nvidia/AceInstruct-1.5B), [AceInstruct-7B](https://huggingface.co/nvidia/AceInstruct-7B), [AceInstruct-72B](https://huggingface.co/nvidia/AceInstruct-72B)


## 連絡先
Zihan Liu (zihanl@nvidia.com), Yang Chen (yachen@nvidia.com), Wei Ping (wping@nvidia.com)


## 引用
本研究が役に立った場合は、引用していただけるとありがたい。
<pre>
@article{acemath2024,
  title={AceMath: Advancing Frontier Math Reasoning with Post-Training and Reward Modeling},
  author={Liu, Zihan and Chen, Yang and Shoeybi, Mohammad and Catanzaro, Bryan and Ping, Wei},
  journal={arXiv preprint},
  year={2024}
}
</pre>


## ライセンス
AceMath-Instruct の学習データセットは非商用利用に限られ、OpenAI が生成したデータの [利用規約](https://openai.com/policies/row-terms-of-use/) に従う。このデータセットは、[Creative Commons Attribution: Non-Commercial 4.0 International](https://spdx.org/licenses/CC-BY-NC-4.0) のライセンスの下で提供する。

<details>
<summary>原文(英語)</summary>

[website](https://research.nvidia.com/labs/adlr/acemath/) | [paper](https://arxiv.org/abs/2412.15084)


## AceMath-Instruct Training Data Card
We release all the datasets to train AceMath-1.5B/7B/72B-Instruct models. These models are built upon the Qwen2.5-Math-Base models through a multi-stage supervised fine-tuning (SFT) process. The fine-tuning begins with general-purpose SFT data (general_sft_stage1.parquet and general_sft_stage2.parquet) and is followed by math-specific SFT data (math_sft.parquet). In our experiments, fine-tuning the Qwen2.5-Math-Base models using only the math-specific SFT data also delivers competitive math reasoning performance.

AceMath-Instruct training datasets are <b>Built with Qwen</b> with math prompt outputs generated by Qwen2.5-Math-72B-Instruct. Outputs for other types of prompts are generated using GPT-4o-mini.

Here are the data statistics:
- general_sft_stage1: 2,261,687 samples (consisting of code & math SFT samples)
- general_sft_stage2: 1,634,573 samples (consisting of code & math & general-domain SFT samples)
- math_sft: 1,661,094 (only math SFT samples)


## Benchmark Results (AceMath-Instruct + AceMath-72B-RM)

<p align="center">
  <img src="./acemath-pic.png" alt="AceMath Benchmark Results" width="800">
</p>


We compare AceMath to leading proprietary and open-access math models in above Table. Our AceMath-7B-Instruct, largely outperforms the previous best-in-class Qwen2.5-Math-7B-Instruct (Average pass@1: 67.2 vs. 62.9) on a variety of math reasoning benchmarks, while coming close to the performance of 10× larger Qwen2.5-Math-72B-Instruct (67.2 vs. 68.2). Notably, our AceMath-72B-Instruct outperforms the state-of-the-art Qwen2.5-Math-72B-Instruct (71.8 vs. 68.2), GPT-4o (67.4) and Claude 3.5 Sonnet (65.6) by a margin. We also report the rm@8 accuracy (best of 8) achieved by our reward model, AceMath-72B-RM, which sets a new record on these reasoning benchmarks. This excludes OpenAI’s o1 model, which relies on scaled inference computation.


## How to use
```python
from datasets import load_dataset

data_files = {
    "general_sft_stage1": "data/general_sft_stage1.parquet",
    "general_sft_stage2": "data/general_sft_stage2.parquet",
    "math_sft": "data/math_sft.parquet",
}

# load the datasets
dataset = load_dataset(
    "nvidia/AceMath-Instruct-Training-Data",
    data_files=data_files,
    cache_dir="CACHE_DIR_TO_STORE_THE_DATASET"
)

# print example
print(dataset['math_sft'][0])

# example format
"""
{
    "messages": [
        {
            "role": "user",
            "content": "...",
        }
    ],
    "answer": "..."
}
"""
```


## All Resources
### AceMath Instruction Models
- [AceMath-1.5B-Instruct](https://huggingface.co/nvidia/AceMath-1.5B-Instruct), [AceMath-7B-Instruct](https://huggingface.co/nvidia/AceMath-7B-Instruct), [AceMath-72B-Instruct](https://huggingface.co/nvidia/AceMath-72B-Instruct)

### AceMath Reward Models
- [AceMath-7B-RM](https://huggingface.co/nvidia/AceMath-7B-RM), [AceMath-72B-RM](https://huggingface.co/nvidia/AceMath-72B-RM)

### Evaluation & Training Data
- [AceMath-RewardBench](https://huggingface.co/datasets/nvidia/AceMath-RewardBench), [AceMath-Instruct Training Data](https://huggingface.co/datasets/nvidia/AceMath-Instruct-Training-Data), [AceMath-RM Training Data](https://huggingface.co/datasets/nvidia/AceMath-RM-Training-Data)

### General Instruction Models
- [AceInstruct-1.5B](https://huggingface.co/nvidia/AceInstruct-1.5B), [AceInstruct-7B](https://huggingface.co/nvidia/AceInstruct-7B), [AceInstruct-72B](https://huggingface.co/nvidia/AceInstruct-72B)


## Correspondence to
Zihan Liu (zihanl@nvidia.com), Yang Chen (yachen@nvidia.com), Wei Ping (wping@nvidia.com)


## Citation
If you find our work helpful, we’d appreciate it if you could cite us.
<pre>
@article{acemath2024,
  title={AceMath: Advancing Frontier Math Reasoning with Post-Training and Reward Modeling},
  author={Liu, Zihan and Chen, Yang and Shoeybi, Mohammad and Catanzaro, Bryan and Ping, Wei},
  journal={arXiv preprint},
  year={2024}
}
</pre>


## License
AceMath-Instruct training dataets are for non-commercial use only, subject to [Terms of Use](https://openai.com/policies/row-terms-of-use/) of the data generated by OpenAI. We put this dataset under the license of [Creative Commons Attribution: Non-Commercial 4.0 International](https://spdx.org/licenses/CC-BY-NC-4.0).

</details>