open-llm-leaderboard/details_Intel__neural-chat-7b-v3

该数据集是在模型Intel/neural-chat-7b-v3在Open LLM Leaderboard上的评估运行期间自动创建的。数据集由64个配置组成，每个配置对应一个被评估的任务。它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分割，使用运行的时间戳命名。train分割始终指向最新结果。一个名为results的额外配置存储了所有运行的聚合结果，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Python中的datasets库加载运行细节的示例。

原始信息汇总

数据集概述

数据集基本信息

• 名称: Evaluation run of Intel/neural-chat-7b-v3
• 来源: 在评估模型 Intel/neural-chat-7b-v3 的过程中自动创建
• 评估平台: Open LLM Leaderboard

数据集结构

• 配置数量: 64个配置
• 每个配置对应: 一个评估任务
• 创建次数: 1次运行
• 分割命名: 使用运行的时间戳命名
• 最新结果: "train" 分割指向最新结果
• 汇总结果: 额外配置 "results" 存储所有运行的汇总结果

数据加载示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_Intel__neural-chat-7b-v3_public", "harness_winogrande_5", split="train")

最新结果

• 来源: 最新结果来自 run 2023-11-14T07:40:49.387630
• 详细结果: python { "all": { "acc": 0.6159787282968994, "acc_stderr": 0.03250548890211772, "acc_norm": 0.6270034478352691, "acc_norm_stderr": 0.03333659536654491, "mc1": 0.42472460220318237, "mc1_stderr": 0.01730400095716748, "mc2": 0.5876875540849609, "mc2_stderr": 0.015599384707939359, "em": 0.43435402684563756, "em_stderr": 0.005076143925092938, "f1": 0.5043477348993302, "f1_stderr": 0.004800980590603821 }, ... }

配置详情

• 配置名称: harness_arc_challenge_25
  • 数据文件:
    • 分割: 2023_11_14T07_40_49.387630
      • 路径: **/details_harness|arc:challenge|25_2023-11-14T07-40-49.387630.parquet
    • 分割: latest
      • 路径: **/details_harness|arc:challenge|25_2023-11-14T07-40-49.387630.parquet
• 配置名称: harness_drop_3
  • 数据文件:
    • 分割: 2023_11_14T07_40_49.387630
      • 路径: **/details_harness|drop|3_2023-11-14T07-40-49.387630.parquet
    • 分割: latest
      • 路径: **/details_harness|drop|3_2023-11-14T07-40-49.387630.parquet
• 配置名称: harness_gsm8k_5
  • 数据文件:
    • 分割: 2023_11_14T07_40_49.387630
      • 路径: **/details_harness|gsm8k|5_2023-11-14T07-40-49.387630.parquet
    • 分割: latest
      • 路径: **/details_harness|gsm8k|5_2023-11-14T07-40-49.387630.parquet
• 配置名称: harness_hellaswag_10
  • 数据文件:
    • 分割: 2023_11_14T07_40_49.387630
      • 路径: **/details_harness|hellaswag|10_2023-11-14T07-40-49.387630.parquet
    • 分割: latest
      • 路径: **/details_harness|hellaswag|10_2023-11-14T07-40-49.387630.parquet
• 配置名称: harness_hendrycksTest_5
  • 数据文件:
    • 分割: 2023_11_14T07_40_49.387630
      • 路径:
        • **/details_harness|hendrycksTest-abstract_algebra|5_2023-11-14T07-40-49.387630.parquet
        • **/details_harness|hendrycksTest-anatomy|5_2023-11-14T07-40-49.387630.parquet
        • **/details_harness|hendrycksTest-astronomy|5_2023-11-14T07-40-49.387630.parquet
        • **/details_harness|hendrycksTest-business_ethics|5_2023-11-14T07-40-49.387630.parquet
        • **/details_harness|hendrycksTest-clinical_knowledge|5_2023-11-14T07-40-49.387630.parquet
        • **/details_harness|hendrycksTest-college_biology|5_2023-11-14T07-40-49.387630.parquet
        • **/details_harness|hendrycksTest-college_chemistry|5_2023-11-14T07-40-49.387630.parquet
        • **/details_harness|hendrycksTest-college_computer_science|5_2023-11-14T07-40-49.387630.parquet
        • **/details_harness|hendrycksTest-college_mathematics|5_2023-11-14T07-40-49.387630.parquet
        • **/details_harness|hendrycksTest-college_medicine|5_2023-11-14T07-40-49.387630.parquet
        • **/details_harness|hendrycksTest-college_physics|5_2023-11-14T07-40-49.387630.parquet
        • **/details_harness|hendrycksTest-computer_security|5_2023-11-14T07-40-49.387630.parquet
        • ...