open-llm-leaderboard-old/details_Eric111__UltraCatunaMayo-DPO

该数据集是在模型Eric111/UltraCatunaMayo-DPO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，每个配置对应一个被评估的任务。数据集是从1次运行中创建的，每次运行在每个配置中表示为特定的分割，train分割始终指向最新的结果。一个名为results的额外配置存储了运行的所有聚合结果，这些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库加载运行中的详细信息的示例。

原始信息汇总

数据集概述

数据集简介

该数据集是在模型 Eric111/UltraCatunaMayo-DPO 的评估运行期间自动创建的，用于 Open LLM Leaderboard。

数据集组成

• 数据集包含 63 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，每次运行可以在每个配置中找到特定的分割，分割名称使用运行的时间戳。
• "train" 分割始终指向最新的结果。
• 额外的 "results" 配置存储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。

数据加载示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_Eric111__UltraCatunaMayo-DPO", "harness_winogrande_5", split="train")

最新结果

以下是 2024-03-24T21:57:21.525992 运行的最新结果：

python { "all": { "acc": 0.6572542102172605, "acc_stderr": 0.03206396348161774, "acc_norm": 0.65708128406132, "acc_norm_stderr": 0.032730675102960426, "mc1": 0.605875152998776, "mc1_stderr": 0.017106588140700325, "mc2": 0.7644277231224181, "mc2_stderr": 0.013925519350259008 }, "harness|arc:challenge|25": { "acc": 0.7064846416382252, "acc_stderr": 0.013307250444941115, "acc_norm": 0.7286689419795221, "acc_norm_stderr": 0.012993807727545803 }, "harness|hellaswag|10": { "acc": 0.7092212706632145, "acc_stderr": 0.0045319353915070065, "acc_norm": 0.8874726150169289, "acc_norm_stderr": 0.0031536835304090366 }, "harness|hendrycksTest-abstract_algebra|5": { "acc": 0.37, "acc_stderr": 0.048523658709391, "acc_norm": 0.37, "acc_norm_stderr": 0.048523658709391 }, # 其他任务的结果省略... }

配置详情

• harness_arc_challenge_25

  • 分割: 2024_03_24T21_57_21.525992
  • 路径: **/details_harness|arc:challenge|25_2024-03-24T21-57-21.525992.parquet
  • 分割: latest
  • 路径: **/details_harness|arc:challenge|25_2024-03-24T21-57-21.525992.parquet

• harness_gsm8k_5

  • 分割: 2024_03_24T21_57_21.525992
  • 路径: **/details_harness|gsm8k|5_2024-03-24T21-57-21.525992.parquet
  • 分割: latest
  • 路径: **/details_harness|gsm8k|5_2024-03-24T21-57-21.525992.parquet

• harness_hellaswag_10

  • 分割: 2024_03_24T21_57_21.525992
  • 路径: **/details_harness|hellaswag|10_2024-03-24T21-57-21.525992.parquet
  • 分割: latest
  • 路径: **/details_harness|hellaswag|10_2024-03-24T21-57-21.525992.parquet

• harness_hendrycksTest_5

  • 分割: 2024_03_24T21_57_21.525992
  • 路径:
    • **/details_harness|hendrycksTest-abstract_algebra|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-anatomy|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-astronomy|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-business_ethics|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-clinical_knowledge|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-college_biology|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-college_chemistry|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-college_computer_science|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-college_mathematics|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-college_medicine|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-college_physics|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-computer_security|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-conceptual_physics|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-econometrics|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-electrical_engineering|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-elementary_mathematics|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-formal_logic|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-global_facts|5_2024-03-24T21-57-21.525992.parquet
    • **/details_harness|hendrycksTest-high_school_biology|5_2024-03-24T21-57-21.525992.parquet
    • `**