answers listlengths 1 1 | ground_truth_ctx dict | document stringclasses 1
value | messages listlengths 1 15 | ctxs listlengths 100 100 |
|---|---|---|---|---|
["क्या आप वर्तमान में अपने भविष्य के लि(...TRUNCATED) | {"ctx":"लाभ नियोजक: उत्तरजीवी | अपने उत्तरज(...TRUNCATED) | ssa | [{"content":"मुझे अपने उत्तराधिकारियों के लि(...TRUNCATED) | [{"text":"लाभ योजनाकार: उत्तरजीवी | अपने उत्(...TRUNCATED) |
["आपको यह सोचना चाहिए कि यदि आपकी अभी मृ(...TRUNCATED) | {"ctx":"लाभ योजनाकार: उत्तरजीवी | अपने उत्त(...TRUNCATED) | ssa | [{"content":"मुझे अपने उत्तराधिकारियों के लि(...TRUNCATED) | [{"text":"लाभ योजनाकार: उत्तरजीवी | अपने उत्(...TRUNCATED) |
["नहीं, आपकी पूर्व पत्नी को मिलने वाले ल(...TRUNCATED) | {"ctx":"आपके जीवित तलाकशुदा जीवनसाथी के लि(...TRUNCATED) | ssa | [{"content":"मुझे अपने उत्तराधिकारियों के लि(...TRUNCATED) | [{"text":"अन्य बातें जो आपको जानने की आवश्यक(...TRUNCATED) |
["उन्हें उत्तरजीवी लाभ मिल सकता है यदि (...TRUNCATED) | {"ctx":"आपके जीवित तलाकशुदा जीवनसाथी के लि(...TRUNCATED) | ssa | [{"content":"मुझे अपने उत्तराधिकारियों के लि(...TRUNCATED) | [{"text":"आपके उत्तरजीवी को कितना मिलेगा आप(...TRUNCATED) |
["यह जानकारी दी गई नहीं है कि पुनर्विवा(...TRUNCATED) | {"ctx":"आपके जीवित तलाकशुदा जीवनसाथी के लि(...TRUNCATED) | ssa | [{"content":"मुझे अपने उत्तराधिकारियों के लि(...TRUNCATED) | [{"text":"अन्य बातें जो आपको जानने की आवश्यक(...TRUNCATED) |
["आपको विकलांग होते ही विकलांगता लाभ क(...TRUNCATED) | {"ctx":"लाभ नियोजक: विकलांगता | आप कैसे आवेद(...TRUNCATED) | ssa | [{"content":"मैं जानना चाहता हूँ कि मुझे विकला(...TRUNCATED) | [{"text":"आपको विकलांग होते ही विकलांगता ला(...TRUNCATED) |
["क्या आपने कानूनी तौर पर अपना नाम बदल ल(...TRUNCATED) | {"ctx":"अमेरिका में जन्मे वयस्क के लिए सही (...TRUNCATED) | ssa | [{"content":"मैं एक सुधारित कार्ड के लिए आवेदन(...TRUNCATED) | [{"text":"गैर-नागरिक वयस्क के लिए संशोधित का(...TRUNCATED) |
["तो सबसे पहले आपको सामाजिक सुरक्षा को (...TRUNCATED) | {"ctx":"अमेरिका में जन्मे वयस्क के लिए सही (...TRUNCATED) | ssa | [{"content":"मैं एक सुधारित कार्ड के लिए आवेदन(...TRUNCATED) | [{"text":"अमेरिका में जन्मे वयस्क के लिए सही(...TRUNCATED) |
["आप अपनी अमेरिकी नागरिकता के प्रमाण क(...TRUNCATED) | {"ctx":"अमेरिका में जन्मे वयस्क के लिए सही (...TRUNCATED) | ssa | [{"content":"मैं एक सुधारित कार्ड के लिए आवेदन(...TRUNCATED) | [{"text":"अमेरिका में जन्मे वयस्क के लिए सही(...TRUNCATED) |
["हम आपसे कोई अन्य दस्तावेज़ मांगेंगे (...TRUNCATED) | {"ctx":"अमेरिका में जन्मे वयस्क के लिए सही (...TRUNCATED) | ssa | [{"content":"मैं एक सुधारित कार्ड के लिए आवेदन(...TRUNCATED) | [{"text":"विदेश में जन्मे अमेरिकी नागरिक वय(...TRUNCATED) |
- データの内容
- データの概要
- 複製について
- 出典
- 論文
- アノテーションの方法
- 微調整に使うときの注意
- 調査メモの記述
- 元のカード(日本語訳)
- データセットの説明:
- データセットの所有者:
- データセットの作成日:
- ライセンス/利用規約:
- 想定される用途:
- データセットの特徴
- データセットの形式
- データ量
- 倫理上の考慮:
- 引用
- Dataset Description:
- Dataset Owner:
- Dataset Creation Date:
- License/Terms of Use:
- Intended Usage:
- Dataset Characterization
- Dataset Format
- Dataset Quantification
- Ethical Considerations:
- Citing
nvidia/ChatRAG-Hi(複製)
データの内容
英語の ChatRAG Bench を GCP で翻訳してヒンディー語にした、文書に基づく会話型 QA のデータ。1 行が 1 件の会話で、発話の並び messages、文書 document、文脈の候補 ctxs、正解の文脈 ground_truth_ctx を持つ。各データセットから約 500 件を、逆翻訳の正確さで選び、質の低い翻訳を取り除いた。ライセンスは CC BY-SA 4.0 である。
データの概要
| 項目 | 値 |
|---|---|
| 行数 | 3,948(8 構成・8 分割の合計) |
| 列数 | 3〜5(構成による) |
| 容量 | 415.40 MB(ファイル 21 本) |
| ファイルの形式 | json 9、py 8、sh 1 |
| 言語 | hi |
| 公開 | public |
構成と分割
| 構成(config) | 分割(split)ごとの行数 | 列数 |
|---|---|---|
doc2dial |
test 498 |
5 |
doqa_cooking |
test 500 |
3 |
doqa_movies |
test 500 |
3 |
doqa_travel |
test 500 |
3 |
hybridial |
test 500 |
3 |
inscit |
test 450 |
5 |
qrecc |
test 500 |
5 |
quac |
test 500 |
5 |
列(構成 doc2dial・分割 test。構成ごとに列が違う)
| 列 | 型 | 先頭の行の値 |
|---|---|---|
answers |
list<string> |
["क्या आप वर्तमान में अपने भविष्य के लि |
ground_truth_ctx |
struct{ctx: string, index: int64} |
{"ctx":"लाभ नियोजक: उत्तरजीवी | अपने उत्तरज |
document |
string |
ssa |
messages |
list<struct{content: string, role: string}> |
[{"content":"मुझे अपने उत्तराधिकारियों के लि |
ctxs |
list<struct{text: string, title: string}> |
[{"text":"लाभ योजनाकार: उत्तरजीवी | अपने उत् |
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/ChatRAG-Hi をリビジョン 8d533f1a5c62eeb614ede1fcf61213e7d5a34282 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/ChatRAG-Hi
- 複製したリビジョン:
8d533f1a5c62eeb614ede1fcf61213e7d5a34282(2026-01-16 11:13:27+00:00) - 元のライセンス表記:
cc-by-sa-4.0
論文
アノテーションの方法
データセットの作成日:
2025 年 4 月
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 (L408) | 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: △ QA/SFT/数学 RM。AceMath は NC |
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。
データセットの説明:
ChatRAG-Hi(Hindi ChatRAG Bench)データセットは、英語版の ChatRAG Bench を元にしている。ChatRAG Bench は、以下の 10 のデータセットからなるとされる: Doc2Dial、QuAC、QReCC、INSCIT、HybriDialogue、DoQA、ConvFinQA。GCP を使って翻訳し、それぞれの集合から約 500 サンプルを、逆翻訳の正確さに基づいて抽出し、質の低い翻訳を取り除いた。
評価の手順は こちら に記載されている。
ほかのヒンディー語のベンチマークデータセット: [ IFEval-Hi | MT-Bench-Hi | GSM8K-Hi | BFCL-Hi | ChatRAG-Hi ]
データセットの所有者:
NVIDIA Corporation
データセットの作成日:
2025 年 4 月
ライセンス/利用規約:
本データセットは Creative Commons Attribution-ShareAlike 4.0 International License(CC BY-SA 4.0)でライセンスされる。
想定される用途:
本データセットは、ヒンディー語で、文書または検索された文脈に基づく大規模言語モデル(LLM)の対話型の質問応答の能力を評価するために使う。
データセットの特徴
データの収集方法
- 合成
ラベル付けの方法
- 合成
データセットの形式
テキスト
データ量
474MB のプロンプトと応答の組からなり、5948 件の個別のサンプルを含む。
倫理上の考慮:
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 の応用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し使用する場合、開発者は社内のモデルチームと協力して、このモデルが該当する業界や用途の要件を満たしているかを確認し、予期しない製品の悪用に対処すべきである。
セキュリティの脆弱性、または NVIDIA の AI に関する懸念は、こちら から報告されたい。
引用
本研究が役立った場合は、論文の引用を検討されたい:
@article{kamath2025benchmarking,
title={Benchmarking Hindi LLMs: A New Suite of Datasets and a Comparative Analysis},
author={Kamath, Anusha and Singla, Kanishk and Paul, Rakesh and Joshi, Raviraj and Vaidya, Utkarsh and Chauhan, Sanjay Singh and Wartikar, Niranjan},
journal={arXiv preprint arXiv:2508.19831},
year={2025}
}
原文(英語)
Dataset Description:
The ChatRAG-Hi (Hindi ChatRAG Bench) dataset is based on the English version of the ChatRAG Bench, which comprises the following ten datasets: Doc2Dial, QuAC, QReCC, INSCIT, HybriDialogue, DoQA, and ConvFinQA. The dataset was translated using GCP, and approximately 500 samples were filtered from each of these sets based on backtranslation accuracy to eliminate poor translations.
The evaluation steps are described here.
Other Hindi benchmark datasets: [ IFEval-Hi | MT-Bench-Hi | GSM8K-Hi | BFCL-Hi | ChatRAG-Hi ]
Dataset Owner:
NVIDIA Corporation
Dataset Creation Date:
April 2025
License/Terms of Use:
This dataset is licensed under the Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0).
Intended Usage:
This dataset is used to evaluate a large language model’s (LLM) conversational QA capability over documents or retrieved context in the Hindi Language.
Dataset Characterization
Data Collection Method
- Synthetic
Labeling Method
- Synthetic
Dataset Format
Text
Dataset Quantification
474MB of prompt-response pairs, comprising 5948 individual samples.
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.
Citing
If you find our work helpful, please consider citing our paper:
@article{kamath2025benchmarking,
title={Benchmarking Hindi LLMs: A New Suite of Datasets and a Comparative Analysis},
author={Kamath, Anusha and Singla, Kanishk and Paul, Rakesh and Joshi, Raviraj and Vaidya, Utkarsh and Chauhan, Sanjay Singh and Wartikar, Niranjan},
journal={arXiv preprint arXiv:2508.19831},
year={2025}
}
- Downloads last month
- 91