five

open-llm-leaderboard-old/details_itsliupeng__llama2_7b_mmlu

收藏
Hugging Face2023-10-25 更新2024-06-22 收录
下载链接:
https://hf-mirror.com/datasets/open-llm-leaderboard-old/details_itsliupeng__llama2_7b_mmlu
下载链接
链接失效反馈
官方服务:
资源简介:
该数据集是在评估模型itsliupeng/llama2_7b_mmlu时自动创建的,包含64个配置,每个配置对应一个评估任务。数据集由2次运行生成,每次运行的结果存储为特定的分割,分割名称使用运行的时间戳。"train"分割始终指向最新的结果。此外,还有一个名为"results"的配置,存储了所有运行的聚合结果,并用于计算和显示在Open LLM Leaderboard上的聚合指标。

该数据集是在评估模型itsliupeng/llama2_7b_mmlu时自动创建的,包含64个配置,每个配置对应一个评估任务。数据集由2次运行生成,每次运行的结果存储为特定的分割,分割名称使用运行的时间戳。"train"分割始终指向最新的结果。此外,还有一个名为"results"的配置,存储了所有运行的聚合结果,并用于计算和显示在Open LLM Leaderboard上的聚合指标。
提供机构:
open-llm-leaderboard-old
原始信息汇总

数据集概述

数据集摘要

该数据集是在评估模型 itsliupeng/llama2_7b_mmluOpen LLM Leaderboard 上的自动创建的。数据集包含 64 个配置,每个配置对应一个评估任务。数据集从 2 次运行中创建,每个运行可以在每个配置中作为一个特定的分片找到,分片名称使用运行的时间戳。"train" 分片总是指向最新的结果。

数据集结构

数据集包含以下配置:

  • harness_arc_challenge_25

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|arc:challenge|25_2023-10-10T15-25-23.413789.parquet
  • harness_drop_3

    • 分片:2023_10_25T10_05_20.920502, latest
    • 路径:**/details_harness|drop|3_2023-10-25T10-05-20.920502.parquet
  • harness_gsm8k_5

    • 分片:2023_10_25T10_05_20.920502, latest
    • 路径:**/details_harness|gsm8k|5_2023-10-25T10-05-20.920502.parquet
  • harness_hellaswag_10

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hellaswag|10_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:多个路径,例如 **/details_harness|hendrycksTest-abstract_algebra|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_abstract_algebra_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-abstract_algebra|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_anatomy_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-anatomy|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_astronomy_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-astronomy|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_business_ethics_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-business_ethics|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_clinical_knowledge_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-clinical_knowledge|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_college_biology_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-college_biology|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_college_chemistry_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-college_chemistry|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_college_computer_science_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-college_computer_science|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_college_mathematics_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-college_mathematics|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_college_medicine_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-college_medicine|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_college_physics_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-college_physics|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_computer_security_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-computer_security|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_conceptual_physics_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-conceptual_physics|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_econometrics_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-econometrics|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_electrical_engineering_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-electrical_engineering|5_2023-10-10T15-25-23.413789.parquet
  • harness_hendrycksTest_elementary_mathematics_5

    • 分片:2023_10_10T15_25_23.413789, latest
    • 路径:**/details_harness|hendrycksTest-elementary_mathematics|5_2023-10-10T15-25-23.413789.parquet

最新结果

最新结果来自运行 2023-10-25T10:05:20.920502,包含以下指标: python { "all": { "em": 0.0012583892617449664, "em_stderr": 0.0003630560893119021, "f1": 0.05594588926174501, "f1_stderr": 0.0013036425627808016, "acc": 0.41156271672651484, "acc_stderr": 0.009842322182656855 }, "harness|drop|3": { "em": 0.0012583892617449664, "em_stderr": 0.0003630560893119021, "f1": 0.05594588926174501, "f1_stderr": 0.0013036425627808016 }, "harness|gsm8k|5": { "acc": 0.07884761182714177, "acc_stderr": 0.00742339051987324 }, "harness|winogrande|5": { "acc": 0.744277821625888, "acc_stderr": 0.012261253845440473 } }

5,000+
优质数据集
54 个
任务类型
进入经典数据集
二维码
社区交流群

面向社区/商业的数据集话题

二维码
科研交流群

面向高校/科研机构的开源数据集话题

数据驱动未来

携手共赢发展

商业合作