open-llm-leaderboard/details_Weyaxi__OpenOrca-Nebula-7B

该数据集是在模型 Weyaxi/OpenOrca-Nebula-7B 在 Open LLM Leaderboard 上的评估运行期间自动创建的。数据集由 61 个配置组成，每个配置对应一个被评估的任务。数据集包含一次运行的结果，每次运行在每个配置中表示为特定的拆分，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 还提供了如何使用 `datasets` 库中的 `load_dataset` 函数加载运行详情的示例。

原始信息汇总

数据集概述

数据集摘要

该数据集是在对模型 Weyaxi/OpenOrca-Nebula-7B 进行评估运行期间自动创建的，用于 Open LLM Leaderboard。数据集包含 61 个配置，每个配置对应一个评估任务。数据集从 1 次运行中创建，每个运行的详细信息可以在每个配置中找到，使用运行的时间戳作为分割名称。"train" 分割始终指向最新的结果。

数据集结构

数据集包含以下配置：

• harness_arc_challenge_25
• harness_hellaswag_10
• harness_hendrycksTest_5

每个配置包含多个分割，例如：

• 2023_11_08T11_58_02.317093
• latest

数据文件路径

每个配置的数据文件路径如下：

• harness_arc_challenge_25:
  • **/details_harness|arc:challenge|25_2023-11-08T11-58-02.317093.parquet
• harness_hellaswag_10:
  • **/details_harness|hellaswag|10_2023-11-08T11-58-02.317093.parquet
• harness_hendrycksTest_5:
  • **/details_harness|hendrycksTest-abstract_algebra|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-anatomy|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-astronomy|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-business_ethics|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-clinical_knowledge|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-college_biology|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-college_chemistry|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-college_computer_science|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-college_mathematics|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-college_medicine|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-college_physics|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-computer_security|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-conceptual_physics|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-econometrics|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-electrical_engineering|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-elementary_mathematics|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-formal_logic|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-global_facts|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-high_school_biology|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-high_school_chemistry|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-high_school_computer_science|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-high_school_european_history|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-high_school_geography|5_2023-11-08T11-58-02.317093.parquet
  • **/details_harness|hendrycksTest-high_school_government_and_politics|5_2023-11-08T11-58-02.317093.parquet

最新结果

最新结果来自 2023-11-08T11:58:02.317093 运行，包含多个任务的评估结果。例如：

• harness|arc:challenge|25:
  • acc: 0.552901023890785
  • acc_norm: 0.5870307167235495
• harness|hellaswag|10:
  • acc: 0.6283608842859988
  • acc_norm: 0.8183628759211312
• harness|hendrycksTest-abstract_algebra|5:
  • acc: 0.32
  • acc_norm: 0.32

数据集加载示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_Weyaxi__OpenOrca-Nebula-7B_public", "harness_truthfulqa_mc_0", split="train")