HYUKI5 commited on
Commit
67496e4
·
verified ·
1 Parent(s): 5e49060

nvidia/OpenMathReasoning@d3d086647557 を複製し、カードを日本語で書く

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +50 -1
  2. README.md +417 -0
  3. data/additional_problems-00000-of-00001.parquet +3 -0
  4. data/cot-00000-of-00144.parquet +3 -0
  5. data/cot-00001-of-00144.parquet +3 -0
  6. data/cot-00002-of-00144.parquet +3 -0
  7. data/cot-00003-of-00144.parquet +3 -0
  8. data/cot-00004-of-00144.parquet +3 -0
  9. data/cot-00005-of-00144.parquet +3 -0
  10. data/cot-00006-of-00144.parquet +3 -0
  11. data/cot-00007-of-00144.parquet +3 -0
  12. data/cot-00008-of-00144.parquet +3 -0
  13. data/cot-00009-of-00144.parquet +3 -0
  14. data/cot-00010-of-00144.parquet +3 -0
  15. data/cot-00011-of-00144.parquet +3 -0
  16. data/cot-00012-of-00144.parquet +3 -0
  17. data/cot-00013-of-00144.parquet +3 -0
  18. data/cot-00014-of-00144.parquet +3 -0
  19. data/cot-00015-of-00144.parquet +3 -0
  20. data/cot-00016-of-00144.parquet +3 -0
  21. data/cot-00017-of-00144.parquet +3 -0
  22. data/cot-00018-of-00144.parquet +3 -0
  23. data/cot-00019-of-00144.parquet +3 -0
  24. data/cot-00020-of-00144.parquet +3 -0
  25. data/cot-00021-of-00144.parquet +3 -0
  26. data/cot-00022-of-00144.parquet +3 -0
  27. data/cot-00023-of-00144.parquet +3 -0
  28. data/cot-00024-of-00144.parquet +3 -0
  29. data/cot-00025-of-00144.parquet +3 -0
  30. data/cot-00026-of-00144.parquet +3 -0
  31. data/cot-00027-of-00144.parquet +3 -0
  32. data/cot-00028-of-00144.parquet +3 -0
  33. data/cot-00029-of-00144.parquet +3 -0
  34. data/cot-00030-of-00144.parquet +3 -0
  35. data/cot-00031-of-00144.parquet +3 -0
  36. data/cot-00032-of-00144.parquet +3 -0
  37. data/cot-00033-of-00144.parquet +3 -0
  38. data/cot-00034-of-00144.parquet +3 -0
  39. data/cot-00035-of-00144.parquet +3 -0
  40. data/cot-00036-of-00144.parquet +3 -0
  41. data/cot-00037-of-00144.parquet +3 -0
  42. data/cot-00038-of-00144.parquet +3 -0
  43. data/cot-00039-of-00144.parquet +3 -0
  44. data/cot-00040-of-00144.parquet +3 -0
  45. data/cot-00041-of-00144.parquet +3 -0
  46. data/cot-00042-of-00144.parquet +3 -0
  47. data/cot-00043-of-00144.parquet +3 -0
  48. data/cot-00044-of-00144.parquet +3 -0
  49. data/cot-00045-of-00144.parquet +3 -0
  50. data/cot-00046-of-00144.parquet +3 -0
.gitattributes CHANGED
@@ -1,6 +1,5 @@
1
  *.7z filter=lfs diff=lfs merge=lfs -text
2
  *.arrow filter=lfs diff=lfs merge=lfs -text
3
- *.avro filter=lfs diff=lfs merge=lfs -text
4
  *.bin filter=lfs diff=lfs merge=lfs -text
5
  *.bz2 filter=lfs diff=lfs merge=lfs -text
6
  *.ckpt filter=lfs diff=lfs merge=lfs -text
@@ -58,3 +57,53 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
58
  # Video files - compressed
59
  *.mp4 filter=lfs diff=lfs merge=lfs -text
60
  *.webm filter=lfs diff=lfs merge=lfs -text
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  *.7z filter=lfs diff=lfs merge=lfs -text
2
  *.arrow filter=lfs diff=lfs merge=lfs -text
 
3
  *.bin filter=lfs diff=lfs merge=lfs -text
4
  *.bz2 filter=lfs diff=lfs merge=lfs -text
5
  *.ckpt filter=lfs diff=lfs merge=lfs -text
 
57
  # Video files - compressed
58
  *.mp4 filter=lfs diff=lfs merge=lfs -text
59
  *.webm filter=lfs diff=lfs merge=lfs -text
60
+ cot/cot-split01.jsonl filter=lfs diff=lfs merge=lfs -text
61
+ cot/cot-split02.jsonl filter=lfs diff=lfs merge=lfs -text
62
+ cot/cot-split03.jsonl filter=lfs diff=lfs merge=lfs -text
63
+ cot/cot-split04.jsonl filter=lfs diff=lfs merge=lfs -text
64
+ cot/cot-split05.jsonl filter=lfs diff=lfs merge=lfs -text
65
+ cot/cot-split06.jsonl filter=lfs diff=lfs merge=lfs -text
66
+ cot/cot-split07.jsonl filter=lfs diff=lfs merge=lfs -text
67
+ cot/cot-split08.jsonl filter=lfs diff=lfs merge=lfs -text
68
+ cot/cot-split09.jsonl filter=lfs diff=lfs merge=lfs -text
69
+ cot/cot-split10.jsonl filter=lfs diff=lfs merge=lfs -text
70
+ cot/cot-split11.jsonl filter=lfs diff=lfs merge=lfs -text
71
+ cot/cot-split12.jsonl filter=lfs diff=lfs merge=lfs -text
72
+ cot/cot-split13.jsonl filter=lfs diff=lfs merge=lfs -text
73
+ cot/cot-split14.jsonl filter=lfs diff=lfs merge=lfs -text
74
+ cot/cot-split15.jsonl filter=lfs diff=lfs merge=lfs -text
75
+ cot/cot-split16.jsonl filter=lfs diff=lfs merge=lfs -text
76
+ cot/cot-split17.jsonl filter=lfs diff=lfs merge=lfs -text
77
+ cot/cot-split18.jsonl filter=lfs diff=lfs merge=lfs -text
78
+ cot/cot-split19.jsonl filter=lfs diff=lfs merge=lfs -text
79
+ cot/cot-split20.jsonl filter=lfs diff=lfs merge=lfs -text
80
+ genselect/genselect-split01.jsonl filter=lfs diff=lfs merge=lfs -text
81
+ genselect/genselect-split02.jsonl filter=lfs diff=lfs merge=lfs -text
82
+ genselect/genselect-split03.jsonl filter=lfs diff=lfs merge=lfs -text
83
+ genselect/genselect-split04.jsonl filter=lfs diff=lfs merge=lfs -text
84
+ genselect/genselect-split05.jsonl filter=lfs diff=lfs merge=lfs -text
85
+ genselect/genselect-split06.jsonl filter=lfs diff=lfs merge=lfs -text
86
+ genselect/genselect-split07.jsonl filter=lfs diff=lfs merge=lfs -text
87
+ genselect/genselect-split08.jsonl filter=lfs diff=lfs merge=lfs -text
88
+ genselect/genselect-split09.jsonl filter=lfs diff=lfs merge=lfs -text
89
+ genselect/genselect-split10.jsonl filter=lfs diff=lfs merge=lfs -text
90
+ tir/tir-split01.jsonl filter=lfs diff=lfs merge=lfs -text
91
+ tir/tir-split02.jsonl filter=lfs diff=lfs merge=lfs -text
92
+ tir/tir-split03.jsonl filter=lfs diff=lfs merge=lfs -text
93
+ tir/tir-split04.jsonl filter=lfs diff=lfs merge=lfs -text
94
+ tir/tir-split05.jsonl filter=lfs diff=lfs merge=lfs -text
95
+ tir/tir-split06.jsonl filter=lfs diff=lfs merge=lfs -text
96
+ tir/tir-split07.jsonl filter=lfs diff=lfs merge=lfs -text
97
+ tir/tir-split08.jsonl filter=lfs diff=lfs merge=lfs -text
98
+ tir/tir-split09.jsonl filter=lfs diff=lfs merge=lfs -text
99
+ tir/tir-split10.jsonl filter=lfs diff=lfs merge=lfs -text
100
+ tir/tir-split11.jsonl filter=lfs diff=lfs merge=lfs -text
101
+ tir/tir-split12.jsonl filter=lfs diff=lfs merge=lfs -text
102
+ tir/tir-split13.jsonl filter=lfs diff=lfs merge=lfs -text
103
+ tir/tir-split14.jsonl filter=lfs diff=lfs merge=lfs -text
104
+ tir/tir-split15.jsonl filter=lfs diff=lfs merge=lfs -text
105
+ tir/tir-split16.jsonl filter=lfs diff=lfs merge=lfs -text
106
+ tir/tir-split17.jsonl filter=lfs diff=lfs merge=lfs -text
107
+ tir/tir-split18.jsonl filter=lfs diff=lfs merge=lfs -text
108
+ tir/tir-split19.jsonl filter=lfs diff=lfs merge=lfs -text
109
+ tir/tir-split20.jsonl filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,417 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ pretty_name: nvidia/OpenMathReasoning (mirror)
4
+ source_datasets:
5
+ - nvidia/OpenMathReasoning
6
+ tags:
7
+ - laya
8
+ - laya-finetune
9
+ - mirror
10
+ - domain-math
11
+ - use-permissive
12
+ - genre-instruction
13
+ - label-text
14
+ - annot-llm
15
+ - lang-en
16
+ language:
17
+ - en
18
+ task_categories:
19
+ - question-answering
20
+ - text-generation
21
+ size_categories:
22
+ - 1M<n<10M
23
+ configs:
24
+ - config_name: default
25
+ data_files:
26
+ - split: cot
27
+ path: data/cot-*
28
+ - split: tir
29
+ path: data/tir-*
30
+ - split: genselect
31
+ path: data/genselect-*
32
+ - split: additional_problems
33
+ path: data/additional_problems-*
34
+ dataset_info:
35
+ features:
36
+ - name: expected_answer
37
+ dtype: string
38
+ - name: problem_type
39
+ dtype: string
40
+ - name: problem_source
41
+ dtype: string
42
+ - name: generation_model
43
+ dtype: string
44
+ - name: pass_rate_72b_tir
45
+ dtype: string
46
+ - name: problem
47
+ dtype: string
48
+ - name: generated_solution
49
+ dtype: string
50
+ - name: inference_mode
51
+ dtype: string
52
+ - name: used_in_kaggle
53
+ dtype: bool
54
+ splits:
55
+ - name: cot
56
+ num_bytes: 71639174648
57
+ num_examples: 3201061
58
+ - name: tir
59
+ num_bytes: 35746562996
60
+ num_examples: 1718466
61
+ - name: genselect
62
+ num_bytes: 6981124435
63
+ num_examples: 565620
64
+ - name: additional_problems
65
+ num_bytes: 66328865
66
+ num_examples: 193170
67
+ download_size: 49585391985
68
+ dataset_size: 114433190944
69
+ ---
70
+
71
+ # nvidia/OpenMathReasoning(複製)
72
+
73
+ ## データの内容
74
+
75
+ 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。
76
+
77
+ ## データの概要
78
+
79
+ | 項目 | 値 |
80
+ |---|---|
81
+ | 行数 | 5,678,317(1 構成・4 分割の合計) |
82
+ | 列数 | 9 |
83
+ | 容量 | 49.53 GB(ファイル 234 本) |
84
+ | ファイルの形式 | parquet 198 |
85
+ | 言語 | en |
86
+ | 公開 | public |
87
+
88
+ ### 構成と分割
89
+
90
+ | 構成(config) | 分割(split)ごとの行数 | 列数 |
91
+ |---|---|---:|
92
+ | `default` | `cot` 3,201,061、`tir` 1,718,466、`genselect` 565,620、`additional_problems` 193,170 | 9 |
93
+
94
+ ### 列(構成 `default`・分割 `cot`)
95
+
96
+ | 列 | 型 | 先頭の行の値 |
97
+ |---|---|---|
98
+ | `expected_answer` | `string` | `\(\frac{C_{n_1}^{a_1} \cdot C_{n_2}^{a_2} \cdot …` |
99
+ | `problem_type` | `string` | `has_answer_extracted` |
100
+ | `problem_source` | `string` | `aops_c6_high_school_olympiads` |
101
+ | `generation_model` | `string` | `DeepSeek-R1` |
102
+ | `pass_rate_72b_tir` | `string` | `0.65625` |
103
+ | `problem` | `string` | `Given a group of \( N \) balls consisting of \( …` |
104
+ | `generated_solution` | `string` | `<think> Okay, so I need to find the probability …` |
105
+ | `inference_mode` | `string` | `cot` |
106
+ | `used_in_kaggle` | `bool` | `True` |
107
+
108
+ 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。
109
+
110
+ ## 複製について
111
+
112
+ [`nvidia/OpenMathReasoning`](https://huggingface.co/datasets/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。
113
+
114
+ ## 出典
115
+
116
+ - 元のデータセット: https://huggingface.co/datasets/nvidia/OpenMathReasoning
117
+ - 複製したリビジョン: `d3d08664755704f422af97d43a7ff0ded4bd95df`(2025-05-27 18:43:44+00:00)
118
+ - 元のライセンス表記: `cc-by-4.0`
119
+
120
+ ## 論文
121
+
122
+ - https://arxiv.org/abs/2504.16891
123
+
124
+ ## アノテーションの方法
125
+
126
+ (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)
127
+
128
+ - **problem**: [AoPS フォーラム](https://artofproblemsolving.com/community) から抽出し、[Qwen2.5-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-32B-Instruct) で整形した問題文。
129
+ - **generated_solution**: [DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1) または [QwQ-32B](https://huggingface.co/Qwen/QwQ-32B) を使って合成により生成された解答。
130
+ - **generation_model**: 生成に使用したモデル。DeepSeek-R1 または QwQ-32B。
131
+ - **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。
132
+ - **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。
133
+ - **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](https://github.com/hendrycks/math) からの生成物も少数含むためである。
134
+ - **inference_mode**: 推論モード。"cot"、"tir"、"genselect" のいずれか。
135
+ - **pass_rate_72b_tir**: TIR モードで実行した [Qwen2.5-Math-72B-Instruct](https://huggingface.co/Qwen/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。
136
+ - **used_in_kaggle**: そのインスタンスが [AIMO-2 Kaggle competition](https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/leaderboard) の優勝モデルの学習に使われたかどうか。[OpenMath-Nemotron-14B-Kaggle](https://huggingface.co/nvidia/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。
137
+
138
+ ## 微調整に使うときの注意
139
+
140
+ - 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。
141
+ - 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。
142
+ - 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。
143
+
144
+ ## 調査メモの記述
145
+
146
+ | メモ | 節 | ライセンス(記載) | 確認 | 注記 |
147
+ |---|---|---|---|---|
148
+ | `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 |
149
+
150
+ ## 元のカード(日本語訳)
151
+
152
+ 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。
153
+
154
+ ---
155
+
156
+ # OpenMathReasoning
157
+
158
+
159
+ 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。
160
+ このデータセットには次が含まれる。
161
+
162
+ * [AoPS フォーラム](https://artofproblemsolving.com/community) から得た 306K 件の一意な数学の問題(次を含む):
163
+ * 3.2M 件の長い chain-of-thought(CoT)解答
164
+ * 1.7M 件の長いツール統合推論(TIR)解答
165
+ * 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)
166
+ * 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)
167
+
168
+ 問題の前処理に [Qwen2.5-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-32B-Instruct) を、解答の生成に [DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1) と [QwQ-32B](https://huggingface.co/Qwen/QwQ-32B) を使用した。
169
+
170
+ このデータセットは、[AIMO-2 Kaggle competition](https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/leaderboard) での我々の優勝提出の基盤となった。
171
+
172
+ 詳細は [論文](https://arxiv.org/abs/2504.16891) を参照のこと。
173
+
174
+ **_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。
175
+
176
+ 問題数のこの食い違いは、2 つの要因で説明される。
177
+ - 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。
178
+ - 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。
179
+
180
+
181
+ **_注記:_** このデー���の初期版は、[Llama-Nemotron-Post-Training-Dataset](https://huggingface.co/datasets/nvidia/Llama-Nemotron-Post-Training-Dataset) で別途公開された。
182
+
183
+ ## データセットのフィールド
184
+
185
+ OpenMathReasoning データセットには、次のフィールドが含まれる:
186
+
187
+ - **problem**: [AoPS フォーラム](https://artofproblemsolving.com/community) から抽出し、[Qwen2.5-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-32B-Instruct) で整形した問題文。
188
+ - **generated_solution**: [DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1) または [QwQ-32B](https://huggingface.co/Qwen/QwQ-32B) を使って合成により生成された解答。
189
+ - **generation_model**: 生成に使用したモデル。DeepSeek-R1 または QwQ-32B。
190
+ - **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。
191
+ - **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。
192
+ - **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](https://github.com/hendrycks/math) からの生成物も少数含むためである。
193
+ - **inference_mode**: 推論モード。"cot"、"tir"、"genselect" のいずれか。
194
+ - **pass_rate_72b_tir**: TIR モードで実行した [Qwen2.5-Math-72B-Instruct](https://huggingface.co/Qwen/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。
195
+ - **used_in_kaggle**: そのインスタンスが [AIMO-2 Kaggle competition](https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/leaderboard) の優勝モデルの学習に使われたかどうか。[OpenMath-Nemotron-14B-Kaggle](https://huggingface.co/nvidia/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。
196
+
197
+ ## OpenMath-Nemotron モデル
198
+
199
+ 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。
200
+
201
+ * [OpenMath-Nemotron-1.5B](https://huggingface.co/nvidia/OpenMath-Nemotron-1.5B)
202
+ * [OpenMath-Nemotron-7B](https://huggingface.co/nvidia/OpenMath-Nemotron-7B)
203
+ * [OpenMath-Nemotron-14B](https://huggingface.co/nvidia/OpenMath-Nemotron-14B)
204
+ * [OpenMath-Nemotron-14B-Kaggle](https://huggingface.co/nvidia/OpenMath-Nemotron-14B-Kaggle)([AIMO-2 Kaggle competition](https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/leaderboard) で使用されたモデル)
205
+ * [OpenMath-Nemotron-32B](https://huggingface.co/nvidia/OpenMath-Nemotron-32B)
206
+
207
+ ![Evaluation Results](./results.png)
208
+
209
+ 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](https://arxiv.org/abs/2504.16891) を参照のこと。
210
+
211
+ | Model | AIME24 | AIME25 | HMMT-24-25 | HLE-Math |
212
+ |-------------------------------|-----------------|-------|-------|-------------|
213
+ | DeepSeek-R1-Distill-Qwen-1.5B | 26.8 (60.0) | 21.4 (36.7) | 14.2 (26.5) | 2.9 (5.0) |
214
+ | [OpenMath-Nemotron-1.5B](https://huggingface.co/nvidia/OpenMath-Nemotron-1.5B) CoT | 61.6 (80.0) | 49.5 (66.7) | 39.9 (53.6) | 5.4 (5.4) |
215
+ | [OpenMath-Nemotron-1.5B](https://huggingface.co/nvidia/OpenMath-Nemotron-1.5B) TIR | 52.0 (83.3) | 39.7 (70.0) | 37.2 (60.7) | 2.5 (6.2) |
216
+ | + 自己 GenSelect | 83.3 | 70.0 | 62.2 | 7.9 |
217
+ | + 32B GenSelect | 83.3 | 70.0 | 62.8 | 8.3 |
218
+ | DeepSeek-R1-Distill-Qwen-7B | 54.4 (80.0) | 38.6 (53.3) | 30.6 (42.9) | 3.3 (5.2) |
219
+ | [OpenMath-Nemotron-7B](https://huggingface.co/nvidia/OpenMath-Nemotron-7B) CoT | 74.8 (80.0) | 61.2 (76.7) | 49.7 (57.7) | 6.6 (6.6) |
220
+ | [OpenMath-Nemotron-7B](https://huggingface.co/nvidia/OpenMath-Nemotron-7B) TIR | 72.9 (83.3) | 57.5 (76.7) | 54.6 (66.3) | 7.8 (10.8) |
221
+ | + 自己 GenSelect | 86.7 | 76.7 | 68.4 | 11.5 |
222
+ | + 32B GenSelect | 86.7 | 76.7 | 69.9 | 11.9 |
223
+ | DeepSeek-R1-Distill-Qwen-14B | 65.8 (80.0) | 48.4 (60.0) | 40.1 (52.0) | 4.2 (4.8) |
224
+ | [OpenMath-Nemotron-14B-MIX (kaggle)](https://huggingface.co/nvidia/OpenMath-Nemotron-14B-Kaggle) | 73.7 (86.7) | 57.9 (73.3) | 50.5 (64.8) | 5.7 (6.5) |
225
+ | [OpenMath-Nemotron-14B](https://huggingface.co/nvidia/OpenMath-Nemotron-14B) CoT | 76.3 (83.3) | 63.0 (76.7) | 52.1 (60.7) | 7.5 (7.6) |
226
+ | [OpenMath-Nemotron-14B](https://huggingface.co/nvidia/OpenMath-Nemotron-14B) TIR | 76.3 (86.7) | 61.3 (76.7) | 58.6 (70.9) | 9.5 (11.5) |
227
+ | + 自己 GenSelect | 86.7 | 76.7 | 72.4 | 14.1 |
228
+ | + 32B GenSelect | 90.0 | 76.7 | 71.9 | 13.7 |
229
+ | QwQ-32B | 78.1 (86.7) | 66.5 (76.7) | 55.9 (63.3) | 9.0 (9.5) |
230
+ | DeepSeek-R1-Distill-Qwen-32B | 66.9 (83.3) | 51.8 (73.3) | 39.9 (51.0) | 4.8 (6.0) |
231
+ | [OpenMath-Nemotron-32B](https://huggingface.co/nvidia/OpenMath-Nemotron-32B) CoT | 76.5 (86.7) | 62.5 (73.3) | 53.0 (59.2) | 8.3 (8.3) |
232
+ | [OpenMath-Nemotron-32B](https://huggingface.co/nvidia/OpenMath-Nemotron-32B) TIR | 78.4 (93.3) | 64.2 (76.7) | 59.7 (70.9) | 9.2 (12.5) |
233
+ | + 自己 GenSelect | 93.3 | 80.0 | 73.5 | 15.7 |
234
+ | DeepSeek-R1 | 79.1 (86.7) | 64.3 (73.3) | 53.0 (59.2) | 10.5 (11.4) |
235
+
236
+ ## 結果の再現
237
+
238
+ 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。
239
+
240
+ - [コード](https://github.com/NVIDIA/NeMo-Skills)
241
+ - [モデル](https://huggingface.co/collections/nvidia/openmathreasoning-68072c0154a5099573d2e730)
242
+ - [データセット](https://huggingface.co/datasets/nvidia/OpenMathReasoning)
243
+ - [論文](https://arxiv.org/abs/2504.16891)
244
+
245
+ 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](https://nvidia.github.io/NeMo-Skills/openmathreasoning1/) を提供する。
246
+
247
+ ## 引用
248
+
249
+ 本研究が役立った場合は、引用を検討すること。
250
+
251
+ ```bibtex
252
+ @article{moshkov2025aimo2,
253
+ title = {AIMO-2 Winning Solution: Building State-of-the-Art Mathematical Reasoning Models with OpenMathReasoning dataset},
254
+ author = {Ivan Moshkov and Darragh Hanley and Ivan Sorokin and Shubham Toshniwal and Christof Henkel and Benedikt Schifferer and Wei Du and Igor Gitman},
255
+ year = {2025},
256
+ journal = {arXiv preprint arXiv:2504.16891}
257
+ }
258
+ ```
259
+
260
+
261
+ ## データセットの所有者:
262
+ NVIDIA Corporation
263
+
264
+ ## 公開日:
265
+ 04/23/2025
266
+
267
+ ## データのバージョン
268
+ 1.0 (04/23/2025)
269
+
270
+ ## ライセンス/利用規約:
271
+ cc-by-4.0
272
+
273
+ ## 想定される用途:
274
+ 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。
275
+
276
+
277
+ ## 倫理的配慮:
278
+ 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。
279
+
280
+ セキュリティの脆弱性や NVIDIA AI に関する懸念は、[こちら](https://www.nvidia.com/en-us/support/submit-security-vulnerability/) から報告すること。
281
+
282
+ <details>
283
+ <summary>原文(英語)</summary>
284
+
285
+ # OpenMathReasoning
286
+
287
+
288
+ OpenMathReasoning is a large-scale math reasoning dataset for training large language models (LLMs).
289
+ This dataset contains
290
+
291
+ * 306K unique mathematical problems sourced from [AoPS forums](https://artofproblemsolving.com/community) with:
292
+ * 3.2M long chain-of-thought (CoT) solutions
293
+ * 1.7M long tool-integrated reasoning (TIR) solutions
294
+ * 566K samples that select the most promising solution out of many candidates (GenSelect)
295
+ * Additional 193K problems sourced from AoPS forums (problems only, no solutions)
296
+
297
+ We used [Qwen2.5-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-32B-Instruct) to preprocess problems, and
298
+ [DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1) and [QwQ-32B](https://huggingface.co/Qwen/QwQ-32B) to generate solutions.
299
+
300
+ This dataset was a foundation of our winning submission to the
301
+ [AIMO-2 Kaggle competition](https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/leaderboard).
302
+
303
+ See our [paper](https://arxiv.org/abs/2504.16891) to learn more details!
304
+
305
+ **_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.
306
+
307
+ Two factors explain this discrepancy in question numbers.
308
+ - 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.
309
+ - 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.
310
+
311
+
312
+
313
+ **_NOTE:_** An early version of this data was released separately in [Llama-Nemotron-Post-Training-Dataset](https://huggingface.co/datasets/nvidia/Llama-Nemotron-Post-Training-Dataset).
314
+
315
+ ## Dataset fields
316
+
317
+ OpenMathReasoning dataset contains the following fields:
318
+
319
+ - **problem**: Problem statement extracted from [AoPS forums](https://artofproblemsolving.com/community) and refined with [Qwen2.5-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-32B-Instruct)
320
+ - **generated_solution**: Synthetically generated solution using either [DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1) or [QwQ-32B](https://huggingface.co/Qwen/QwQ-32B)
321
+ - **generation_model**: DeepSeek-R1 or QwQ-32B
322
+ - **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.
323
+ - **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.
324
+ - **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](https://github.com/hendrycks/math).
325
+ - **inference_mode**: "cot", "tir" or "genselect"
326
+ - **pass_rate_72b_tir**: Pass rate out of 32 generations for [Qwen2.5-Math-72B-Instruct](https://huggingface.co/Qwen/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.
327
+ - **used_in_kaggle**: Whether the instance was used in training the winning model for [AIMO-2 Kaggle competition](https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/leaderboard) or not. We had used 2.2M CoT and 15K TIR solutions for training the [OpenMath-Nemotron-14B-Kaggle](https://huggingface.co/nvidia/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.
328
+
329
+ ## OpenMath-Nemotron models
330
+
331
+ To demonstrate the quality of this dataset, we release a series of OpenMath-Nemotron models trained on this data.
332
+
333
+ * [OpenMath-Nemotron-1.5B](https://huggingface.co/nvidia/OpenMath-Nemotron-1.5B)
334
+ * [OpenMath-Nemotron-7B](https://huggingface.co/nvidia/OpenMath-Nemotron-7B)
335
+ * [OpenMath-Nemotron-14B](https://huggingface.co/nvidia/OpenMath-Nemotron-14B)
336
+ * [OpenMath-Nemotron-14B-Kaggle](https://huggingface.co/nvidia/OpenMath-Nemotron-14B-Kaggle) (this is the model used in [AIMO-2 Kaggle competition](https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/leaderboard))
337
+ * [OpenMath-Nemotron-32B](https://huggingface.co/nvidia/OpenMath-Nemotron-32B)
338
+
339
+ ![Evaluation Results](./results.png)
340
+
341
+ The models achieve state-of-the-art results on popular mathematical benchmarks. We present metrics as pass@1 (maj@64) where pass@1
342
+ is an average accuracy across 64 generations and maj@64 is the result of majority voting.
343
+ Please see our [paper](https://arxiv.org/abs/2504.16891) for more details on the evaluation setup.
344
+
345
+ | Model | AIME24 | AIME25 | HMMT-24-25 | HLE-Math |
346
+ |-------------------------------|-----------------|-------|-------|-------------|
347
+ | DeepSeek-R1-Distill-Qwen-1.5B | 26.8 (60.0) | 21.4 (36.7) | 14.2 (26.5) | 2.9 (5.0) |
348
+ | [OpenMath-Nemotron-1.5B](https://huggingface.co/nvidia/OpenMath-Nemotron-1.5B) CoT | 61.6 (80.0) | 49.5 (66.7) | 39.9 (53.6) | 5.4 (5.4) |
349
+ | [OpenMath-Nemotron-1.5B](https://huggingface.co/nvidia/OpenMath-Nemotron-1.5B) TIR | 52.0 (83.3) | 39.7 (70.0) | 37.2 (60.7) | 2.5 (6.2) |
350
+ | + Self GenSelect | 83.3 | 70.0 | 62.2 | 7.9 |
351
+ | + 32B GenSelect | 83.3 | 70.0 | 62.8 | 8.3 |
352
+ | DeepSeek-R1-Distill-Qwen-7B | 54.4 (80.0) | 38.6 (53.3) | 30.6 (42.9) | 3.3 (5.2) |
353
+ | [OpenMath-Nemotron-7B](https://huggingface.co/nvidia/OpenMath-Nemotron-7B) CoT | 74.8 (80.0) | 61.2 (76.7) | 49.7 (57.7) | 6.6 (6.6) |
354
+ | [OpenMath-Nemotron-7B](https://huggingface.co/nvidia/OpenMath-Nemotron-7B) TIR | 72.9 (83.3) | 57.5 (76.7) | 54.6 (66.3) | 7.8 (10.8) |
355
+ | + Self GenSelect | 86.7 | 76.7 | 68.4 | 11.5 |
356
+ | + 32B GenSelect | 86.7 | 76.7 | 69.9 | 11.9 |
357
+ | DeepSeek-R1-Distill-Qwen-14B | 65.8 (80.0) | 48.4 (60.0) | 40.1 (52.0) | 4.2 (4.8) |
358
+ | [OpenMath-Nemotron-14B-MIX (kaggle)](https://huggingface.co/nvidia/OpenMath-Nemotron-14B-Kaggle) | 73.7 (86.7) | 57.9 (73.3) | 50.5 (64.8) | 5.7 (6.5) |
359
+ | [OpenMath-Nemotron-14B](https://huggingface.co/nvidia/OpenMath-Nemotron-14B) CoT | 76.3 (83.3) | 63.0 (76.7) | 52.1 (60.7) | 7.5 (7.6) |
360
+ | [OpenMath-Nemotron-14B](https://huggingface.co/nvidia/OpenMath-Nemotron-14B) TIR | 76.3 (86.7) | 61.3 (76.7) | 58.6 (70.9) | 9.5 (11.5) |
361
+ | + Self GenSelect | 86.7 | 76.7 | 72.4 | 14.1 |
362
+ | + 32B GenSelect | 90.0 | 76.7 | 71.9 | 13.7 |
363
+ | QwQ-32B | 78.1 (86.7) | 66.5 (76.7) | 55.9 (63.3) | 9.0 (9.5) |
364
+ | DeepSeek-R1-Distill-Qwen-32B | 66.9 (83.3) | 51.8 (73.3) | 39.9 (51.0) | 4.8 (6.0) |
365
+ | [OpenMath-Nemotron-32B](https://huggingface.co/nvidia/OpenMath-Nemotron-32B) CoT | 76.5 (86.7) | 62.5 (73.3) | 53.0 (59.2) | 8.3 (8.3) |
366
+ | [OpenMath-Nemotron-32B](https://huggingface.co/nvidia/OpenMath-Nemotron-32B) TIR | 78.4 (93.3) | 64.2 (76.7) | 59.7 (70.9) | 9.2 (12.5) |
367
+ | + Self GenSelect | 93.3 | 80.0 | 73.5 | 15.7 |
368
+ | DeepSeek-R1 | 79.1 (86.7) | 64.3 (73.3) | 53.0 (59.2) | 10.5 (11.4) |
369
+
370
+ ## Reproducing our results
371
+
372
+ The pipeline we used to produce the data and models is fully open-sourced!
373
+
374
+ - [Code](https://github.com/NVIDIA/NeMo-Skills)
375
+ - [Models](https://huggingface.co/collections/nvidia/openmathreasoning-68072c0154a5099573d2e730)
376
+ - [Dataset](https://huggingface.co/datasets/nvidia/OpenMathReasoning)
377
+ - [Paper](https://arxiv.org/abs/2504.16891)
378
+
379
+ We provide [all instructions](https://nvidia.github.io/NeMo-Skills/openmathreasoning1/)
380
+ to fully reproduce our results, including data generation.
381
+
382
+ ## Citation
383
+
384
+ If you find our work useful, please consider citing us!
385
+
386
+ ```bibtex
387
+ @article{moshkov2025aimo2,
388
+ title = {AIMO-2 Winning Solution: Building State-of-the-Art Mathematical Reasoning Models with OpenMathReasoning dataset},
389
+ author = {Ivan Moshkov and Darragh Hanley and Ivan Sorokin and Shubham Toshniwal and Christof Henkel and Benedikt Schifferer and Wei Du and Igor Gitman},
390
+ year = {2025},
391
+ journal = {arXiv preprint arXiv:2504.16891}
392
+ }
393
+ ```
394
+
395
+
396
+ ## Dataset Owner(s):
397
+ NVIDIA Corporation
398
+
399
+ ## Release Date:
400
+ 04/23/2025
401
+
402
+ ## Data Version
403
+ 1.0 (04/23/2025)
404
+
405
+ ## License/Terms of Use:
406
+ cc-by-4.0
407
+
408
+ ## Intended Usage:
409
+ 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.
410
+
411
+
412
+ ## Ethical Considerations:
413
+ 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.
414
+
415
+ Please report security vulnerabilities or NVIDIA AI Concerns [here](https://www.nvidia.com/en-us/support/submit-security-vulnerability/).
416
+
417
+ </details>
data/additional_problems-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c5bd7f69efe5d4718f76cd3d5a5e1e6a9f5398c48cfddb0ad64663b2bc428e5
3
+ size 26105718
data/cot-00000-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6640e85f89bb829702a7d30b622acbe2bdb76bb943eb37d77ef20ab145081ddc
3
+ size 220414147
data/cot-00001-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:11a74afa30a2acfaee333d217268c47d6f8792a9df404f1dd718db4265636533
3
+ size 221998437
data/cot-00002-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3a2f4fe017d06e7d73cf86cd4bcd88fec6377430874803cc035cbcc6eb15dfc6
3
+ size 221647098
data/cot-00003-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd478737420d2cc72c43d969dc943b671072a702013bc7dd79007284183de3f9
3
+ size 222820474
data/cot-00004-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a17a7a6746c552afacd86d1bd4276075e4804afabf0dedcce4454a79cdc4ac5
3
+ size 219968119
data/cot-00005-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29e8c61e1961319c16a3ee0e4c024b77c48b232244d92d1e6250884fbe258d7a
3
+ size 221131152
data/cot-00006-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d928c49adae45904a7be0f6b2b407891e86248bbaeef116b498f6634966c4d1
3
+ size 222204796
data/cot-00007-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c4fbca22e4387871a718e357e0e643331b8d3aa8a6e3cc80ad379a75d7ae46c
3
+ size 221268578
data/cot-00008-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c35d496adadc1afd4a45c5e37e484fd8bd2021aa2917ab086b54f11d38179f4a
3
+ size 220744509
data/cot-00009-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58b8c9a2e19ef737049331ef5f9d18845cd84fccd45a707214792ccd7160730f
3
+ size 219921373
data/cot-00010-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c65fae986656e2cbca2746fd2c8341ff5788cc7087b324f23ff2fadbeb3d067a
3
+ size 221180937
data/cot-00011-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a66b554c4fa156caaec73d856eb4d77ad9d387822a1df55459c96fd64e67774e
3
+ size 220512727
data/cot-00012-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4245f6c869ca0632d75585c3414dfdf5208871c359962a6ebf7cddad58b199f
3
+ size 220218705
data/cot-00013-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5de6d3f11b22c0dfaed647332d3fd0222f33bc6be48dc2fbc71bb57aa5bc32b9
3
+ size 220316648
data/cot-00014-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2849be84886ab6a95608828ae8562c42468ff4a8408da28d51ec803a501857e1
3
+ size 221651823
data/cot-00015-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2317eeb27e0d8060678f64e60eee6233b0dfb5ac37539f16900d69fbd5604e65
3
+ size 221463205
data/cot-00016-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:618e15c77a237ea77f3eb5eacaccc89062548eb9eaac607663e4aa13e8f3b588
3
+ size 221775445
data/cot-00017-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a56eea15af569cdb093453f932e97ae656e8f0c945206226a1eda280f99a6e51
3
+ size 221799886
data/cot-00018-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f09e8e42e903268eb89bd00eef71f0f04d0daa7c62203d73188ac495d00957fb
3
+ size 222206536
data/cot-00019-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8c7e0a0d19041d036f41b22eb1a414875a31108e342df0e88d8c23433d16c0f9
3
+ size 219814522
data/cot-00020-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92c4ea2ebdc6cbd1a156c554e6e34426d4aa2654aa77b422dd216683072cef38
3
+ size 222891123
data/cot-00021-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d03f6216ac78a70f6a182bc61a9325a1bd65851f3ef915afd96a3d7f33260f5b
3
+ size 219907961
data/cot-00022-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c523d54d5a1772b65dc493f68a387e55aaf0ab8c6f2c521165df89a628317c65
3
+ size 221761655
data/cot-00023-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d9dc744af2432148e8badd648e690ca5b53999acd7346cf4ba886c3769084dd8
3
+ size 220352968
data/cot-00024-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:540e602947b2db8e49c2437b9d545a431a15830bff1556925dd23c8d72f7afcb
3
+ size 221082471
data/cot-00025-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e5c43cdb5ddd2481f29639ad9518a38f1a0ca3d9e007253db16b655a0b00087
3
+ size 221844181
data/cot-00026-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26015a38bb05dc2359426f17e2324b4818903b3b89390f445b1daae744236497
3
+ size 220846067
data/cot-00027-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b808cefbc99c756c357dacc55a51663b18ec484d58243827016ae3bca9cbcaf
3
+ size 220932741
data/cot-00028-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31e379cd314ac3a022a5b84709b5a4699cf626161c5f2bf5702d71904aaccf2a
3
+ size 221296922
data/cot-00029-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cc66e2c91b7741e3b9a1166ddd4779d8576012f2ce476151ab9bffeb8f2b5edc
3
+ size 221668215
data/cot-00030-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d76f6e6c6e01c9d73987335b3479e5b21dea55b70917015c0ea773c0e8cf988
3
+ size 220293194
data/cot-00031-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:53192e41fab71c74ceaea5c2fbbc2f0e755895377bbbe80772a880bf5c4fda84
3
+ size 221301694
data/cot-00032-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe8af1f6a13007e3541721cea70623b7b6820b509e5aa469ec0bdf86afef9cfc
3
+ size 222326330
data/cot-00033-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b47707cddde932df77cba32f9f9723a0e68ff0dec34238f81274778bd377724
3
+ size 221781208
data/cot-00034-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:045f3312ee0eb8b459af3b0ba9f6f366049fa1b8a62e920fae5379944ddf2e07
3
+ size 221528458
data/cot-00035-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c72f6970a7b300fbfece95ca21c4f6ede1f0b8d92bf19849b7944a82899c08b8
3
+ size 221741329
data/cot-00036-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b857209239f6f44c30a6faf8698869c661659fa8c976d87dfedbef916d76987
3
+ size 221144463
data/cot-00037-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b48309ce3a9a39635ffd2924999d1dc960a96ce8d8ce8c1fdc51f160f1ad9d8b
3
+ size 221095167
data/cot-00038-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:632b6e371100ec53af82093d42d3358369e38e4dcd9b22f3db507002bd3b42f2
3
+ size 220726377
data/cot-00039-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a23d9e589d46a0be017436eb5300ed4ad9bc3c2f73ddacbbc8fcaaac78fa31a1
3
+ size 221897460
data/cot-00040-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51cdc65abaa51fd4f371d15e1cfd2080ae6353b65b4095593e2e1275fd1d1053
3
+ size 220890195
data/cot-00041-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c5303770390a6846b3b29354ff360c2e3b53d1d52c2fdbbbd9e8599840924b7
3
+ size 221296013
data/cot-00042-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ef50aa7913dc2b1185e244c8b0235b5238de3e59b4008cb834c3fef5f88c4295
3
+ size 221545671
data/cot-00043-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a8c4978c5e42b5125bfaf2d98d1c02c1edc87c0ae02d11703627eda2011631c
3
+ size 221670934
data/cot-00044-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd2661a4abc209bf627865db3e426cf5323bd7307df6257467aa729f6a200119
3
+ size 221280949
data/cot-00045-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98bbfd4b1275a9b8866664c813c8a96ce05c9ee9006d26aa7ad2aec9c734caa9
3
+ size 220905201
data/cot-00046-of-00144.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6eda64f728b372bc31e72c7c7212066f15b0a00d133eb4933bd34d5d70451556
3
+ size 221950060