open-llm-leaderboard-old/details_wandb__zephyr-orpo-7b-v0.2

该数据集是在模型 wandb/zephyr-orpo-7b-v0.2 在 Open LLM Leaderboard 上的评估运行期间自动创建的。数据集由 63 个配置组成，每个配置对应一个评估任务。数据集来自一次运行，每次运行在每个配置中表示为特定的拆分，使用运行的时间戳命名。train 拆分始终指向最新结果。一个名为 results 的额外配置存储了所有运行的聚合结果，用于计算和显示 Open LLM Leaderboard 上的聚合指标。README 还提供了加载数据集详细信息的 Python 代码片段，并列出了特定运行的最新结果。

原始信息汇总

数据集概述

数据集名称

• 名称: Evaluation run of wandb/zephyr-orpo-7b-v0.2

数据集组成

• 配置数量: 63个配置，每个配置对应一个评估任务。
• 数据来源: 数据集由1次运行创建，每个运行结果作为特定分割存储，分割名称使用运行的时间戳。
• 最新结果: "train"分割始终指向最新结果。
• 汇总结果: 额外的"results"配置存储所有运行的汇总结果，用于计算和显示在Open LLM Leaderboard上的汇总指标。

数据加载示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_wandb__zephyr-orpo-7b-v0.2", "harness_winogrande_5", split="train")

最新结果

• 最新结果时间戳: 2024-04-15T15:46:25.433441
• 结果摘要: python { "all": { "acc": 0.6338014240730799, "acc_stderr": 0.0323862119285172, "acc_norm": 0.639307612845627, "acc_norm_stderr": 0.033040847206224806, "mc1": 0.2962056303549572, "mc1_stderr": 0.015983595101811392, "mc2": 0.4552587654708832, "mc2_stderr": 0.014573126803250782 }, "harness|arc:challenge|25": { "acc": 0.591296928327645, "acc_stderr": 0.014365750345427001, "acc_norm": 0.6168941979522184, "acc_norm_stderr": 0.014206472661672876 }, "harness|hellaswag|10": { "acc": 0.6443935471021709, "acc_stderr": 0.0047771835089498145, "acc_norm": 0.845947022505477, "acc_norm_stderr": 0.0036026174466413886 }, "harness|hendrycksTest-abstract_algebra|5": { "acc": 0.28, "acc_stderr": 0.04512608598542128, "acc_norm": 0.28, "acc_norm_stderr": 0.04512608598542128 }, # 其他任务的结果... }

配置详情

• 配置名称: harness_arc_challenge_25

  • 数据文件:
    • 分割: 2024_04_15T15_46_25.433441
      • 路径: **/details_harness|arc:challenge|25_2024-04-15T15-46-25.433441.parquet
    • 分割: latest
      • 路径: **/details_harness|arc:challenge|25_2024-04-15T15-46-25.433441.parquet

• 配置名称: harness_gsm8k_5

  • 数据文件:
    • 分割: 2024_04_15T15_46_25.433441
      • 路径: **/details_harness|gsm8k|5_2024-04-15T15-46-25.433441.parquet
    • 分割: latest
      • 路径: **/details_harness|gsm8k|5_2024-04-15T15-46-25.433441.parquet

• 配置名称: harness_hellaswag_10

  • 数据文件:
    • 分割: 2024_04_15T15_46_25.433441
      • 路径: **/details_harness|hellaswag|10_2024-04-15T15-46-25.433441.parquet
    • 分割: latest
      • 路径: **/details_harness|hellaswag|10_2024-04-15T15-46-25.433441.parquet

• 配置名称: harness_hendrycksTest_5

  • 数据文件:
    • 分割: 2024_04_15T15_46_25.433441
      • 路径:
        • **/details_harness|hendrycksTest-abstract_algebra|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-anatomy|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-astronomy|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-business_ethics|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-clinical_knowledge|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-college_biology|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-college_chemistry|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-college_computer_science|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-college_mathematics|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-college_medicine|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-college_physics|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-computer_security|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-conceptual_physics|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-econometrics|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-electrical_engineering|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-elementary_mathematics|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-formal_logic|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-global_facts|5_2024-04-15T15-46-25.433441.parquet
        • **/details_harness|hendrycksTest-high_school_biology|5_2024-04-15T15-46-25.433441.parquet

        • 其他文件路径...

    • 分割: latest
      • 路径:
        • **/details_harness|hendrycksTest-abstract_algebra|5_2024-04-15T15-46-25.433441.parquet

        • 其他文件路径...