Dataset Viewer
Auto-converted to Parquet Duplicate
question_id
int64
0
13k
question
stringlengths
3
7
corpus_id
stringclasses
1 value
pos_doc
listlengths
1
1
neg_doc
listlengths
4
4
0
q_2249
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15954" }, { "id": "d_15945" }, { "id": "d_15955" }, { "id": "d_15950" } ]
1
q_15895
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15954" } ]
2
q_12328
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15954" }, { "id": "d_15950" }, { "id": "d_15955" }, { "id": "d_15939" } ]
3
q_8272
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15955" }, { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15954" } ]
4
q_1991
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15954" }, { "id": "d_15950" } ]
5
q_273
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15950" }, { "id": "d_15953" }, { "id": "d_15939" }, { "id": "d_15954" } ]
6
q_9714
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15945" }, { "id": "d_15953" }, { "id": "d_15939" }, { "id": "d_15954" } ]
7
q_13502
EmotionClassification
[ { "id": "d_15945" } ]
[ { "id": "d_15939" }, { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15950" } ]
8
q_5121
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15939" }, { "id": "d_15950" } ]
9
q_12676
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15939" }, { "id": "d_15945" }, { "id": "d_15955" }, { "id": "d_15950" } ]
10
q_12859
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15945" }, { "id": "d_15955" }, { "id": "d_15953" }, { "id": "d_15939" } ]
11
q_4165
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15945" }, { "id": "d_15950" } ]
12
q_8306
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15950" }, { "id": "d_15945" } ]
13
q_6078
EmotionClassification
[ { "id": "d_15950" } ]
[ { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15955" }, { "id": "d_15939" } ]
14
q_15754
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15954" }, { "id": "d_15955" }, { "id": "d_15939" }, { "id": "d_15950" } ]
15
q_14570
EmotionClassification
[ { "id": "d_15939" } ]
[ { "id": "d_15953" }, { "id": "d_15950" }, { "id": "d_15955" }, { "id": "d_15945" } ]
16
q_11932
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15955" }, { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15950" } ]
17
q_2115
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15950" }, { "id": "d_15945" } ]
18
q_8730
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15945" }, { "id": "d_15953" }, { "id": "d_15950" }, { "id": "d_15939" } ]
19
q_14378
EmotionClassification
[ { "id": "d_15950" } ]
[ { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15955" } ]
20
q_6563
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15955" }, { "id": "d_15953" } ]
21
q_5529
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15939" }, { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15945" } ]
22
q_14654
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15945" }, { "id": "d_15950" } ]
23
q_11245
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15939" }, { "id": "d_15955" }, { "id": "d_15950" }, { "id": "d_15945" } ]
24
q_3818
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15950" }, { "id": "d_15954" }, { "id": "d_15945" }, { "id": "d_15939" } ]
25
q_13623
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15953" }, { "id": "d_15939" }, { "id": "d_15954" }, { "id": "d_15945" } ]
26
q_2982
EmotionClassification
[ { "id": "d_15950" } ]
[ { "id": "d_15955" }, { "id": "d_15939" }, { "id": "d_15953" }, { "id": "d_15945" } ]
27
q_7482
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15945" }, { "id": "d_15954" }, { "id": "d_15950" }, { "id": "d_15939" } ]
28
q_12207
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15954" }, { "id": "d_15950" }, { "id": "d_15939" }, { "id": "d_15955" } ]
29
q_1832
EmotionClassification
[ { "id": "d_15950" } ]
[ { "id": "d_15954" }, { "id": "d_15955" }, { "id": "d_15945" }, { "id": "d_15953" } ]
30
q_13864
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15945" }, { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15950" } ]
31
q_1633
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15950" } ]
32
q_5777
EmotionClassification
[ { "id": "d_15945" } ]
[ { "id": "d_15955" }, { "id": "d_15939" }, { "id": "d_15954" }, { "id": "d_15953" } ]
33
q_144
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15950" }, { "id": "d_15955" } ]
34
q_13931
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15953" }, { "id": "d_15939" }, { "id": "d_15954" }, { "id": "d_15950" } ]
35
q_11373
EmotionClassification
[ { "id": "d_15945" } ]
[ { "id": "d_15939" }, { "id": "d_15950" }, { "id": "d_15954" }, { "id": "d_15953" } ]
36
q_8373
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15950" }, { "id": "d_15953" }, { "id": "d_15955" }, { "id": "d_15945" } ]
37
q_7627
EmotionClassification
[ { "id": "d_15950" } ]
[ { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15954" } ]
38
q_5695
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15955" }, { "id": "d_15954" } ]
39
q_253
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15953" }, { "id": "d_15939" }, { "id": "d_15945" }, { "id": "d_15954" } ]
40
q_7929
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15945" }, { "id": "d_15950" }, { "id": "d_15953" }, { "id": "d_15955" } ]
41
q_13938
EmotionClassification
[ { "id": "d_15939" } ]
[ { "id": "d_15945" }, { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15953" } ]
42
q_3359
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15953" }, { "id": "d_15939" }, { "id": "d_15950" }, { "id": "d_15945" } ]
43
q_2418
EmotionClassification
[ { "id": "d_15939" } ]
[ { "id": "d_15954" }, { "id": "d_15953" }, { "id": "d_15955" }, { "id": "d_15945" } ]
44
q_4959
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15939" }, { "id": "d_15954" }, { "id": "d_15945" }, { "id": "d_15953" } ]
45
q_2677
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15939" }, { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15954" } ]
46
q_6252
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15955" }, { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15954" } ]
47
q_4373
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15955" }, { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15939" } ]
48
q_1769
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15939" }, { "id": "d_15950" }, { "id": "d_15955" }, { "id": "d_15953" } ]
49
q_3905
EmotionClassification
[ { "id": "d_15950" } ]
[ { "id": "d_15954" }, { "id": "d_15939" }, { "id": "d_15955" }, { "id": "d_15953" } ]
50
q_6558
EmotionClassification
[ { "id": "d_15939" } ]
[ { "id": "d_15953" }, { "id": "d_15954" }, { "id": "d_15950" }, { "id": "d_15955" } ]
51
q_15856
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15953" }, { "id": "d_15950" }, { "id": "d_15955" }, { "id": "d_15945" } ]
52
q_9855
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15950" }, { "id": "d_15939" }, { "id": "d_15954" }, { "id": "d_15945" } ]
53
q_1381
EmotionClassification
[ { "id": "d_15939" } ]
[ { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15955" }, { "id": "d_15954" } ]
54
q_8547
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15945" }, { "id": "d_15950" }, { "id": "d_15939" }, { "id": "d_15955" } ]
55
q_13202
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15953" }, { "id": "d_15939" }, { "id": "d_15954" }, { "id": "d_15945" } ]
56
q_1729
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15945" }, { "id": "d_15950" }, { "id": "d_15955" }, { "id": "d_15939" } ]
57
q_1799
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15954" } ]
58
q_10420
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15939" }, { "id": "d_15945" }, { "id": "d_15954" }, { "id": "d_15950" } ]
59
q_7219
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15945" }, { "id": "d_15950" }, { "id": "d_15939" }, { "id": "d_15955" } ]
60
q_7038
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15945" }, { "id": "d_15954" }, { "id": "d_15950" }, { "id": "d_15955" } ]
61
q_7993
EmotionClassification
[ { "id": "d_15945" } ]
[ { "id": "d_15950" }, { "id": "d_15953" }, { "id": "d_15939" }, { "id": "d_15955" } ]
62
q_15611
EmotionClassification
[ { "id": "d_15950" } ]
[ { "id": "d_15939" }, { "id": "d_15953" }, { "id": "d_15955" }, { "id": "d_15945" } ]
63
q_4803
EmotionClassification
[ { "id": "d_15945" } ]
[ { "id": "d_15954" }, { "id": "d_15953" }, { "id": "d_15950" }, { "id": "d_15955" } ]
64
q_2590
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15950" }, { "id": "d_15953" }, { "id": "d_15939" }, { "id": "d_15954" } ]
65
q_15927
EmotionClassification
[ { "id": "d_15950" } ]
[ { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15945" }, { "id": "d_15939" } ]
66
q_1901
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15950" }, { "id": "d_15939" }, { "id": "d_15954" }, { "id": "d_15955" } ]
67
q_9050
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15945" }, { "id": "d_15954" }, { "id": "d_15953" }, { "id": "d_15939" } ]
68
q_14285
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15953" }, { "id": "d_15954" }, { "id": "d_15939" }, { "id": "d_15950" } ]
69
q_5402
EmotionClassification
[ { "id": "d_15950" } ]
[ { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15939" }, { "id": "d_15945" } ]
70
q_12383
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15939" }, { "id": "d_15950" }, { "id": "d_15955" }, { "id": "d_15953" } ]
71
q_4965
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15945" }, { "id": "d_15950" }, { "id": "d_15954" }, { "id": "d_15953" } ]
72
q_8402
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15939" }, { "id": "d_15950" }, { "id": "d_15945" }, { "id": "d_15955" } ]
73
q_4840
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15954" }, { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15955" } ]
74
q_13789
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15950" }, { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15955" } ]
75
q_10607
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15945" }, { "id": "d_15954" }, { "id": "d_15955" }, { "id": "d_15950" } ]
76
q_7098
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15955" }, { "id": "d_15950" }, { "id": "d_15939" }, { "id": "d_15954" } ]
77
q_4304
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15953" }, { "id": "d_15954" }, { "id": "d_15950" }, { "id": "d_15939" } ]
78
q_4446
EmotionClassification
[ { "id": "d_15939" } ]
[ { "id": "d_15945" }, { "id": "d_15953" }, { "id": "d_15954" }, { "id": "d_15955" } ]
79
q_4398
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15955" }, { "id": "d_15954" }, { "id": "d_15950" }, { "id": "d_15939" } ]
80
q_10434
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15954" }, { "id": "d_15950" } ]
81
q_12533
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15954" }, { "id": "d_15955" }, { "id": "d_15945" }, { "id": "d_15939" } ]
82
q_8763
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15939" }, { "id": "d_15950" }, { "id": "d_15945" }, { "id": "d_15954" } ]
83
q_9135
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15945" }, { "id": "d_15950" }, { "id": "d_15955" }, { "id": "d_15954" } ]
84
q_125
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15954" }, { "id": "d_15939" }, { "id": "d_15955" }, { "id": "d_15950" } ]
85
q_8498
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15954" }, { "id": "d_15939" }, { "id": "d_15950" }, { "id": "d_15955" } ]
86
q_12660
EmotionClassification
[ { "id": "d_15939" } ]
[ { "id": "d_15953" }, { "id": "d_15955" }, { "id": "d_15950" }, { "id": "d_15954" } ]
87
q_9118
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15950" }, { "id": "d_15945" }, { "id": "d_15955" }, { "id": "d_15939" } ]
88
q_11838
EmotionClassification
[ { "id": "d_15950" } ]
[ { "id": "d_15953" }, { "id": "d_15954" }, { "id": "d_15955" }, { "id": "d_15939" } ]
89
q_8174
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15939" }, { "id": "d_15955" }, { "id": "d_15953" }, { "id": "d_15950" } ]
90
q_9302
EmotionClassification
[ { "id": "d_15950" } ]
[ { "id": "d_15939" }, { "id": "d_15955" }, { "id": "d_15953" }, { "id": "d_15945" } ]
91
q_9370
EmotionClassification
[ { "id": "d_15954" } ]
[ { "id": "d_15939" }, { "id": "d_15950" }, { "id": "d_15945" }, { "id": "d_15955" } ]
92
q_15503
EmotionClassification
[ { "id": "d_15953" } ]
[ { "id": "d_15950" }, { "id": "d_15954" }, { "id": "d_15939" }, { "id": "d_15955" } ]
93
q_13498
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15954" }, { "id": "d_15945" }, { "id": "d_15939" }, { "id": "d_15950" } ]
94
q_14616
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15939" }, { "id": "d_15954" }, { "id": "d_15953" }, { "id": "d_15950" } ]
95
q_3462
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15954" }, { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15950" } ]
96
q_4941
EmotionClassification
[ { "id": "d_15939" } ]
[ { "id": "d_15954" }, { "id": "d_15950" }, { "id": "d_15953" }, { "id": "d_15945" } ]
97
q_1099
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15950" }, { "id": "d_15953" }, { "id": "d_15939" }, { "id": "d_15954" } ]
98
q_15275
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15945" }, { "id": "d_15953" }, { "id": "d_15939" }, { "id": "d_15950" } ]
99
q_4569
EmotionClassification
[ { "id": "d_15955" } ]
[ { "id": "d_15939" }, { "id": "d_15953" }, { "id": "d_15945" }, { "id": "d_15954" } ]
End of preview. Expand in Data Studio

nvidia/embed-nemotron-dataset-v1(複製)

データの内容

埋め込みモデル llama-embed-nemotron-8b の学習に使われた、微調整用データの選抜サブセット。MIRACL、HotpotQA、MS MARCO などの公開データと、分類タスク用の合成データからなる 26 の構成がある。1 件はクエリ(question)と正例・負例の文書(pos_doc、neg_doc)の組で、一部は文書の本文を持たず ID だけを提供し、取得用のスクリプトが付く。研究開発専用で、FEVER と MIRACL は CC BY-SA 4.0、合成分類データには Llama の利用条件が及ぶ可能性がある。

データの概要

項目 値
行数 12,294,089(26 構成・26 分割の合計)
列数 1〜5(構成による)
容量 2.30 GB(ファイル 69 本)
ファイルの形式 parquet 28、json 25
言語 multilingual
公開 public

構成と分割

構成(config) 分割(split)ごとの行数 列数
EmotionClassification train 13,039 5
EmotionClassification_corpus train 6 1
FEVER train 140,085 5
FEVER_corpus train 235,948 2
GooAQ train 100,000 5
GooAQ_corpus train 86,393 1
HotpotQA_corpus train 754,756 1
MAmmoTH2 train 317,180 5
MIRACL train 79,648 5
MIRACL_corpus train 2,571,803 2
MSMARCO train 532,751 5
MSMARCO_corpus train 3,672,883 1
NFCorpus train 3,685 5
NFCorpus_corpus train 3,573 1
NaturalQuestions train 100,231 5
NaturalQuestions_corpus train 75,215 1
PAQ train 1,000,000 5
PAQ_corpus train 932,307 1
SQuAD train 87,599 5
SQuAD_corpus train 18,891 1

(ほか 6 構成。全体は HF のデータセットビューアで見られる)

列(構成 EmotionClassification・分割 train。構成ごとに列が違う)

列 型 先頭の行の値
question_id int64 0
question string q_2249
corpus_id string EmotionClassification
pos_doc list<struct{id: string}> [1 件]
neg_doc list<struct{id: string}> [4 件]

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/embed-nemotron-dataset-v1 をリビジョン 5003f42aac552e6fc1ed4ed0f5009cf3a70d080f 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

データセットの作成日

2025 年 10 月 21 日


データの収集方法

ハイブリッド: 合成、自動、人手

ラベル付けの方法

ハイブリッド: 合成、自動、人手


微調整に使うときの注意

  • 調査メモの台帳での区分: B(継承条件・gated・モデル由来の条件・上流の規約などが付く)。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 (L317) cc-by-sa-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: △ 埋め込み学習用(検索ペア)。SyntheticClassificationData 104 万問は Llama 条件。ラベルは正例・負例文書

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。


Embed Nemotron Dataset V1

バージョン

Date Commit Changes
2026-01-05 8808454 初版

データセットの説明

このデータセットは、NVIDIA が公開した llama-embed-nemotron-8b モデルを支える、高品質な微調整用データセットを集めたものである。

llama-embed-nemotron-8b は、NVIDIA が学習させた汎用のテキスト埋め込みモデルであり、検索、リランキング、意味的類似度、分類の用途に最適化されている。多言語および言語横断のテキスト検索に強い能力を持ち、テキストベースの検索拡張生成(RAG)システムの基盤となる部品として設計されている。このモデルは、2025 年 10 月 21 日時点で、多言語の MTEB リーダーボードにおいて最先端の性能を達成している。

embed-nemotron-dataset-v1 は、llama-embed-nemotron-8b モデルの学習に使われた微調整用データの選抜サブセットである。このデータの組み合わせは、幅広いタスクで頑健性と高い性能を確保するために、多様な出所から意図的に選別している。内容は、MIRACL、HotpotQA、MS MARCO などの確立された公開データセットから取った非合成データと、タスクの種類の多様性を高めるために特別に生成した合成データ(たとえば分類)の両方からなる。各データセットには、学習の信号を最大化するための高度なハードネガティブマイニングの処理も加えてある。詳細は技術報告を参照: Llama-Embed-Nemotron-8B: A Universal Text Embedding Model for Multilingual and Cross-Lingual Tasks。

下の表に、公開しているすべてのサブデータセットを詳細とともに示す。データを直接再配布しているものもある。残りについては、クエリと文書の ID だけを提供している。後者の場合は、提供された ID から元のテキストを取得するための Python データ処理スクリプト を提供している。

このデータセットは研究開発専用である。


データセットの所有者

NVIDIA Corporation


データセットの作成日

2025 年 10 月 21 日


ライセンス/利用規約

データセットの適用条件: このコレクションの各データセットの利用は、それに付属するライセンスに従う。ただし FEVER と MIRACL のデータセットは Creative Commons Attribution-ShareAlike 4.0 International License(CC BY-SA 4.0)に従い、SyntheticClassificationData は Creative Commons Attribution 4.0 International License(CC BY 4.0)に従う。

追加情報: SyntheticClassificationData データセットには、次のモデルで作成した合成データが含まれる: meta-llama/Llama-3.3-70B-Instruct、meta-llama/Llama-4-Scout-17B-16E-Instruct、meta-llama/Llama-4-Maverick-17B-128E-Instruct。このデータセットを使って、配布または公開される AI モデルを作成・学習・微調整・改良する場合、そのモデルには Llama 3.3 Community License Agreement および Llama 4 Community License Agreement の再配布と利用に関する要件が適用される可能性がある。


想定される用途

embed-nemotron-dataset-v1 は、コミュニティがオープンなモデルの改良を続けるために使われることを想定している。データは学習と評価に自由に使ってよい。


データセットの構成

Dataset Name Task Samples (Queries / Documents) Size (MB) Governing Terms Source
EmotionClassification 分類 13,039 / 6 0.2 公開
FEVER 検索 140,085 / 235,948 118.2 CC BY-SA 4.0 公開
GooAQ 検索 100,000 / 86,393 5.5 公開
HotpotQA 検索 170,000 / 754,756 21.4 公開
MAmmoTH2 検索 317,180 / 317,205 27.8 公開
MIRACL 検索 79,648 / 2,571,803 1,480.0 CC BY-SA 4.0 公開
MSMARCO 検索 532,751 / 3,672,883 80.6 公開
NFCorpus 検索 3,685 / 3,573 0.2 公開
NaturalQuestions 検索 100,231 / 75,215 5.1 公開
PAQ 検索 1,000,000 / 932,307 108.0 公開
SQuAD 検索 87,599 / 18,891 3.4 公開
SciFact 検索 919 / 3,255 0.1 公開
SyntheticClassificationData 分類 1,044,212 / 382,227 440.1 CC BY 4.0 合成
TriviaQA 検索 73,346 / 64,137 3.8 公開
合計(14) 3,662,695 / 9,118,599 2,314.4

タグの凡例

  • 公開: データの出所が別の公開データセットであること
  • 合成: データが合成的に生成されたこと

データセットの数量

  • データセット数の合計: 14
  • サンプル数の合計: 3,662,695 クエリ; 9,118,599 文書
  • 合計サイズ: 2.3 GB

データセットの特徴付け

データの収集方法

ハイブリッド: 合成、自動、人手

ラベル付けの方法

ハイブリッド: 合成、自動、人手


データセットの形式

各データセットは 2 つの部分からなる:

  1. クエリ
    • question_id — クエリの ID
    • question — クエリのテキスト(テキストを直接再配布していない場合はクエリの ID)
    • corpus_id — データセットの名前
    • pos_doc — そのクエリの正例文書のリスト
    • neg_doc — そのクエリの負例文書のリスト
  2. コーパス
    • id — クエリの pos_doc と neg_doc のリストに対応する文書の ID
    • 任意の text — テキストを直接再配布している場合の文書のテキスト

使い方

このデータは NeMo AutoModel フレームワークで使えるように整えてある。

このデータセットは、NVIDIA が公開した llama-embed-nemotron-8b モデルを支える、高品質な微調整用データセットを集めたものである。データを準備してモデルの学習を始めるには、ここ に示した手順に従う。


引用

@misc{babakhin2025llamaembednemotron8buniversaltextembedding,
      title={Llama-Embed-Nemotron-8B: A Universal Text Embedding Model for Multilingual and Cross-Lingual Tasks}, 
      author={Yauhen Babakhin and Radek Osmulski and Ronay Ak and Gabriel Moreira and Mengyao Xu and Benedikt Schifferer and Bo Liu and Even Oldridge},
      year={2025},
      eprint={2511.07025},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2511.07025}, 

倫理的な配慮

NVIDIA は、信頼できる AI は共同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実践を定めている。 利用規約に従ってダウンロードまたは利用する場合、開発者は社内のモデルチームと協力し、このデータセットが該当する業界や用途の要件を満たしていること、また予期しない製品の誤用に対処していることを確認すべきである。

品質、リスク、セキュリティ上の脆弱性、または NVIDIA AI Concerns は こちら から報告してほしい。

原文(英語)

Embed Nemotron Dataset V1

Versions

Date Commit Changes
2026-01-05 8808454 Initial Release

Dataset Description

This dataset is a compilation of high quality fine-tuning datasets that support NVIDIA's release of llama-embed-nemotron-8b model.

llama-embed-nemotron-8b is a versatile text embedding model trained by NVIDIA and optimized for retrieval, reranking, semantic similarity, and classification use cases. This model has robust capabilities for multilingual and cross-lingual text retrieval and is designed to serve as a foundational component in text-based Retrieval-Augmented Generation (RAG) systems. This model achieves state-of-the-art performance on the multilingual MTEB leaderboard as of October 21, 2025.

embed-nemotron-dataset-v1 is a selected subset of the fine-tuning data used for training the llama-embed-nemotron-8b model. This data mix is intentionally curated from a variety of sources to ensure robustness and high performance across a broad range of tasks. It comprises both non-synthetic data drawn from established public datasets like MIRACL, HotpotQA, MS MARCO etc.; and synthetic data specifically generated to enhance diversity in task types (for example, classification). Each dataset is additionally transformed with a sophisticated hard negative mining process to maximize the learning signal. Read more details in our technical report: Llama-Embed-Nemotron-8B: A Universal Text Embedding Model for Multilingual and Cross-Lingual Tasks.

In the table below, you can see all the subdatasets that we are publishing, with their details. In some cases, the data is redistributed directly. For the remaining ones, we only provide IDs of the queries and documents. In the latter case, we provide python data processing script to fetch original texts given provided IDs.

This dataset is for research and development only.


Dataset Owner

NVIDIA Corporation


Dataset Creation Date

October 21, 2025


License/Terms of Use

Dataset Governing Terms: Use of each dataset in this collection is governed by the license accompanying it, except for the FEVER and MIRACL datasets, which are governed by the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0), and the SyntheticClassificationData, which is governed by the Creative Commons Attribution 4.0 International License (CC BY 4.0).

Dataset Additional Information: The SyntheticClassificationData dataset contains synthetic data created using the following models: meta-llama/Llama-3.3-70B-Instruct, meta-llama/Llama-4-Scout-17B-16E-Instruct and meta-llama/Llama-4-Maverick-17B-128E-Instruct. If this dataset is used to create, train, fine tune, or otherwise improve an AI model, which is distributed or made available, such AI model may be subject to redistribution and use requirements in the Llama 3.3 Community License Agreement and Llama 4 Community License Agreement.


Intended Usage

The embed-nemotron-dataset-v1 is intended to be used by the community to continue to improve open models. The data may be freely used to train and evaluate.


Dataset Composition

Dataset Name Task Samples (Queries / Documents) Size (MB) Governing Terms Source
EmotionClassification Classification 13,039 / 6 0.2 public
FEVER Retrieval 140,085 / 235,948 118.2 CC BY-SA 4.0 public
GooAQ Retrieval 100,000 / 86,393 5.5 public
HotpotQA Retrieval 170,000 / 754,756 21.4 public
MAmmoTH2 Retrieval 317,180 / 317,205 27.8 public
MIRACL Retrieval 79,648 / 2,571,803 1,480.0 CC BY-SA 4.0 public
MSMARCO Retrieval 532,751 / 3,672,883 80.6 public
NFCorpus Retrieval 3,685 / 3,573 0.2 public
NaturalQuestions Retrieval 100,231 / 75,215 5.1 public
PAQ Retrieval 1,000,000 / 932,307 108.0 public
SQuAD Retrieval 87,599 / 18,891 3.4 public
SciFact Retrieval 919 / 3,255 0.1 public
SyntheticClassificationData Classification 1,044,212 / 382,227 440.1 CC BY 4.0 synthetic
TriviaQA Retrieval 73,346 / 64,137 3.8 public
Total (14) 3,662,695 / 9,118,599 2,314.4

Tag Legend

  • public: Origin of the data is another public dataset
  • synthetic: The data was synthetically generated

Dataset Quantification

  • Total Number of Datasets: 14
  • Total Number of Samples: 3,662,695 Queries; 9,118,599 Documents
  • Total Size: 2.3 GB

Dataset Characterization

Data Collection Method

Hybrid: Synthetic, Automated, Human

Labeling Method

Hybrid: Synthetic, Automated, Human


Dataset Format

Each given dataset includes two parts:

  1. Queries
    • question_id -- id of the query
    • question -- query text (or query ID if texts are not redistributed directly)
    • corpus_id -- name of the dataset
    • pos_doc -- list of the positive documents for the given query
    • neg_doc -- list of the negative documents for the given query
  2. Corpus
    • id -- document id corresponding to pos_doc and neg_doc lists from the Queries
    • Optional text -- document text if texts are redistributed directly

Usage

The data has been prepared to be used with NeMo AutoModel framework.

This dataset is a compilation of high quality fine-tuning datasets that support NVIDIA's release of llama-embed-nemotron-8b model. To prepare the data and start training the model, follow the steps provided here.


Citation:

@misc{babakhin2025llamaembednemotron8buniversaltextembedding,
      title={Llama-Embed-Nemotron-8B: A Universal Text Embedding Model for Multilingual and Cross-Lingual Tasks}, 
      author={Yauhen Babakhin and Radek Osmulski and Ronay Ak and Gabriel Moreira and Mengyao Xu and Benedikt Schifferer and Bo Liu and Even Oldridge},
      year={2025},
      eprint={2511.07025},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2511.07025}, 

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 dataset meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please report quality, risk, security vulnerabilities or NVIDIA AI Concerns here.

Downloads last month
73

Paper for HYUKI5/nvidia__embed-nemotron-dataset-v1