HYUKI5's picture
nvidia/LiveCodeBench-CPP@6c21015d3095 を複製し、カードを日本語で書く
e5f83f7 verified
|
Raw History Blame Contribute Delete
13.3 kB
metadata
license: cc-by-4.0
pretty_name: nvidia/LiveCodeBench-CPP (mirror)
source_datasets:
  - nvidia/LiveCodeBench-CPP
tags:
  - laya
  - laya-finetune
  - mirror
  - domain-code
  - use-permissive
  - genre-exam
  - label-text
  - annot-natural
size_categories:
  - n<1K
configs:
  - config_name: test
    data_files:
      - split: v5_2408_2501
        path: test_v5_2408_2501.jsonl
      - split: v6_2408_2505
        path: test_v6_2408_2505.jsonl
dataset_info:
  - config_name: test
    features:
      - name: question_title
        dtype: string
      - name: question_content
        dtype: string
      - name: platform
        dtype: string
      - name: question_id
        dtype: string
      - name: contest_id
        dtype: string
      - name: contest_date
        dtype: timestamp[s]
      - name: starter_code
        dtype: string
      - name: difficulty
        dtype: string
      - name: public_test_cases
        dtype: string
      - name: private_test_cases
        dtype: string
      - name: metadata
        dtype: string
    splits:
      - name: v5_2408_2501
        num_bytes: 1802637484
        num_examples: 279
      - name: v6_2408_2505
        num_bytes: 1939402000
        num_examples: 454

nvidia/LiveCodeBench-CPP(複製)

データの内容

LiveCodeBench の release_v6 から、2024 年 10 月〜2025 年 5 月の AtCoder と LeetCode の問題を集め、C++ の解答で評価するためのデータ。1 行が 1 題で、問題文 question_content、プラットフォーム platform、難易度 difficulty、公開と非公開のテストケース public_test_cases・private_test_cases を持つ。AtCoder の問題は標準入力から読み標準出力へ書く形式、LeetCode の問題はスターターコードの関数を呼び出す形式で、C++ の解答をコンパイルして採点する。

データの概要

項目 値
行数 733(1 構成・2 分割の合計)
列数 11
容量 3.74 GB(ファイル 4 本)
ファイルの形式 jsonl 2
公開 public

構成と分割

構成(config) 分割(split)ごとの行数 列数
test v5_2408_2501 279、v6_2408_2505 454 11

列(構成 test)

列 型
question_title string
question_content string
platform string
question_id string
contest_id string
contest_date timestamp[s]
starter_code string
difficulty string
public_test_cases string
private_test_cases string
metadata string

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/LiveCodeBench-CPP をリビジョン 6c21015d3095e5c45812950d683f5f3e65403c13 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

  • 元のカードに論文へのリンクは見つからなかった

アノテーションの方法

データセットの作成日

2025 年 4 月〜5 月

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 (L404) 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。


LiveCodeBench-CPP: C++ でデータ汚染のない評価を行うための LiveCodeBench の拡張

Overview

LiveCodeBench-CPP は、LiveCodeBench の release_v6 から 454 問を含む。期間は 2024 年 10 月から 2025 年 5 月までである。問題の出典は AtCoder(287 問)と LeetCode(167 問)である。

  • AtCoder の問題: 生成された解答は、標準入力(stdin)から入力を読み、標準出力(stdout)へ出力しなければならない。単体テストでは、生成された C++ の解答をコンパイルし、stdin で与えられるテストケースで実行する。
  • LeetCode の問題: 解答は、スターターコードで与えられる既定の関数シグネチャを使って書かなければならない。単体テストでは、生成された C++ の解答をコンパイルし、対象の関数に入力のパラメータを直接渡して呼び出し、返された値を期待される出力と比較する。

LiveCodeBench の詳細は https://huggingface.co/datasets/livecodebench/code_generation_lite を参照。

このデータセットは、商用・非商用のどちらの用途にも使える。

Data Distribution

問題の出典は https://huggingface.co/datasets/livecodebench/code_generation_lite である。

Source # Sample
AtCoder 287
LeetCode 167
Total 合計

Data Fields

Field Type Description
question_title string プログラミング問題のタイトル
question_content string プログラミング問題の本文
platform string AtCoder か LeetCode のいずれかを示す
question_id string 問題ごとに一意な ID
contest_id string コンテストごとに一意な ID
contest_date string コンテストの日付
starter_code string 問題に与えられるスターターコード
difficulty string 難易度を示すラベル(easy、medium、hard のいずれか)
public_test_cases string 公開テストケースの JSON 文字列
private_test_cases string 非公開テストケースの JSON 文字列
meta_data string 空であるか、"func_name" を含む dict オブジェクト

使い方

from datasets import load_dataset
dataset = load_dataset("nvidia/LiveCodeBench-CPP", split="v6_2408_2505")

データ量

  • レコード数: 454 サンプル。
  • ダウンロードサイズ: 1.9GB

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〜5 月

データセットの特徴づけ

** データ収集の方法

  • [ハイブリッド: 自動・合成]

** ラベル付けの方法

  • [ハイブリッド: 自動・合成]

ライセンス・利用条件

準拠する条件: このデータセットは、Creative Commons Attribution 4.0 International License(CC BY 4.0)のもとで提供される。ライセンス本文は https://creativecommons.org/licenses/by/4.0/legalcode にある。

データの開発者: NVIDIA

用途:

コード生成とコードの批評に特化させるため、大規模言語モデル(LLM)を訓練する開発者向け。

公開日:

05/11/2025

データのバージョン

1.0 (05/11/2025)

想定される利用方法

LiveCodeBench-CPP データセットは、コミュニティが公開モデルを改良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、モデルの評価に自由に使ってよい。ただし、利用者はデータセットのライセンスが目的に合うかどうかを自ら確認する責任がある。

倫理上の配慮:

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 応用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て、このモデルが関連する業界と用途の要件を満たすこと、そして想定外の製品の誤用に対処することを確かめるべきである。

セキュリティ上の脆弱性や NVIDIA AI に関する懸念は、こちらで報告を受け付けている。

原文(英語)

LiveCodeBench-CPP: An Extension of LiveCodeBench for Contamination Free Evaluation in C++

Overview

LiveCodeBench-CPP includes 454 problems from the release_v6 of LiveCodeBench, covering the period from October 2024 to May 2025. These problems are sourced from AtCoder (287 problems) and LeetCode (167 problems).

  • AtCoder Problems: These require generated solutions to read inputs from standard input (stdin) and write outputs to standard output (stdout). For unit testing, the generated C++ solutions are compiled and executed using test cases provided via stdin.
  • LeetCode Problems: These require solutions to be written using a predefined function signature provided in a starter code. For unit testing, the generated C++ solutions are compiled and the target function is invoked with input parameters directly, then the returned values are compared against the expected outputs.

To learn about LiveCodeBench, see details at https://huggingface.co/datasets/livecodebench/code_generation_lite.

This dataset is ready for commercial/non-commercial use.

Data Distribution

The problems are sourced from https://huggingface.co/datasets/livecodebench/code_generation_lite.

Source # Sample
AtCoder 287
LeetCode 167
Total 454

Data Fields

Field Type Description
question_title string A title for the programming question
question_content string The content of the programming question
platform string Indicates either AtCoder or LeetCode
question_id string A unique id for each question
contest_id string A unique id for the contest
contest_date string Contest date
starter_code string A starter code for the given question
difficulty string A label indicate difficulty (easy, medium, or hard)
public_test_cases string A JSON string for the public test cases
private_test_cases string A JSON string for the private test cases
meta_data string A dict object (either empty or include "func_name")

How to Use It

from datasets import load_dataset
dataset = load_dataset("nvidia/LiveCodeBench-CPP", split="v6_2408_2505")

Dataset Quantification

  • Record Count - 454 samples.
  • Download Size - 1.9GB

Dataset Owner(s)

NVIDIA Corporation

Dataset Creation Date

April 2025 - May 2025

Dataset Characterization

** Data Collection Method

  • [Hybrid: Automated, Synthetic]

** Labeling Method

  • [Hybrid: Automated, Synthetic]

License/Terms of Use

GOVERNING TERMS: This dataset is licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0) available at https://creativecommons.org/licenses/by/4.0/legalcode.

Data Developer: NVIDIA

Use Case:

Developers training Large Language Models (LLMs) to specialize LLMs in code generation and code critique.

Release Date:

05/11/2025

Data Version

1.0 (05/11/2025)

Intended Usage

The LiveCodeBench-CPP Dataset is intended to be used by the community to continue to improve open models. The data may be freely used to evaluate models. However, the user is responsible for checking if the dataset license is fit for the intended purpose.

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.