open-llm-leaderboard-old/details_jan-hq__trinity-v1

该数据集是在Open LLM Leaderboard上对模型jan-hq/trinity-v1进行评估时自动创建的。它由63个配置组成，每个配置对应一个评估任务。数据集是从1次运行中生成的，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。一个额外的配置results存储了所有运行的聚合结果，这些结果用于计算和显示Open LLM Leaderboard上的聚合指标。可以使用`datasets`库中的`load_dataset`函数加载该数据集。

原始信息汇总

数据集概述

数据集基本信息

• 名称: Evaluation run of jan-hq/trinity-v1
• 来源: 自动创建于模型 jan-hq/trinity-v1 在 Open LLM Leaderboard 的评估运行中。
• 组成: 包含 63 个配置，每个配置对应一个评估任务。
• 创建次数: 从 1 次运行中创建。每个运行在每个配置中作为一个特定的分割存在，分割名称使用运行的时间戳。"train" 分割始终指向最新结果。
• 额外配置: "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_jan-hq__trinity-v1", "harness_winogrande_5", split="train")

最新结果

• 最新结果时间戳: 2023-12-16T19:24:08.553660
• 详细结果: python { "all": { "acc": 0.6575877329335247, "acc_stderr": 0.031985421208388404, "acc_norm": 0.6571647268300141, "acc_norm_stderr": 0.032648337921958155, "mc1": 0.5507955936352509, "mc1_stderr": 0.01741294198611529, "mc2": 0.6931209356367747, "mc2_stderr": 0.015031530031665238 }, "harness|arc:challenge|25": { "acc": 0.6988054607508533, "acc_stderr": 0.013406741767847632, "acc_norm": 0.7226962457337884, "acc_norm_stderr": 0.013082095839059376 }, "harness|hellaswag|10": { "acc": 0.711113324039036, "acc_stderr": 0.004523188431142894, "acc_norm": 0.8835889265086636, "acc_norm_stderr": 0.0032006176493464752 }, "harness|hendrycksTest-abstract_algebra|5": { "acc": 0.33, "acc_stderr": 0.04725815626252605, "acc_norm": 0.33, "acc_norm_stderr": 0.04725815626252605 }, # 其他任务结果省略... }

配置信息

• 配置名称: harness_arc_challenge_25

  • 分割: 2023_12_16T19_24_08.553660
    • 路径: **/details_harness|arc:challenge|25_2023-12-16T19-24-08.553660.parquet
  • 分割: latest
    • 路径: **/details_harness|arc:challenge|25_2023-12-16T19-24-08.553660.parquet

• 配置名称: harness_gsm8k_5

  • 分割: 2023_12_16T19_24_08.553660
    • 路径: **/details_harness|gsm8k|5_2023-12-16T19-24-08.553660.parquet
  • 分割: latest
    • 路径: **/details_harness|gsm8k|5_2023-12-16T19-24-08.553660.parquet

• 配置名称: harness_hellaswag_10

  • 分割: 2023_12_16T19_24_08.553660
    • 路径: **/details_harness|hellaswag|10_2023-12-16T19-24-08.553660.parquet
  • 分割: latest
    • 路径: **/details_harness|hellaswag|10_2023-12-16T19-24-08.553660.parquet

• 配置名称: harness_hendrycksTest_5

  • 分割: 2023_12_16T19_24_08.553660
    • 路径:
      • **/details_harness|hendrycksTest-abstract_algebra|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-anatomy|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-astronomy|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-business_ethics|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-clinical_knowledge|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-college_biology|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-college_chemistry|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-college_computer_science|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-college_mathematics|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-college_medicine|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-college_physics|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-computer_security|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-conceptual_physics|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-econometrics|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-electrical_engineering|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-elementary_mathematics|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-formal_logic|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-global_facts|5_2023-12-16T19-24-08.553660.parquet
      • **/details_harness|hendrycksTest-high_school_biology|5_2023-12-16T19-24-08.553660.parquet

      其他路径省略...

  }