five

open-llm-leaderboard/details_uukuguy__speechless-orca-platypus-coig-lite-4k-0.5e-13b

收藏
Hugging Face2023-10-18 更新2024-03-04 收录
下载链接:
https://hf-mirror.com/datasets/open-llm-leaderboard/details_uukuguy__speechless-orca-platypus-coig-lite-4k-0.5e-13b
下载链接
链接失效反馈
官方服务:
资源简介:
该数据集是在Open LLM Leaderboard上对模型uukuguy/speechless-orca-platypus-coig-lite-4k-0.5e-13b进行评估时自动创建的。数据集由64个配置组成,每个配置对应一个评估任务。数据集由2次运行创建,每次运行的结果作为特定配置中的一个分割,分割名称使用运行的时间戳。train分割始终指向最新结果。此外,还有一个名为results的配置,存储所有运行的聚合结果,并用于计算和显示Open LLM Leaderboard上的聚合指标。
提供机构:
open-llm-leaderboard
原始信息汇总

数据集概述

该数据集是在对模型 uukuguy/speechless-orca-platypus-coig-lite-4k-0.5e-13b 进行评估运行期间自动创建的。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" 分片始终指向最新的结果。

数据集结构

配置详情

  • harness_arc_challenge_25

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|arc:challenge|25_2023-08-30T23:37:31.114358.parquet
  • harness_drop_3

    • 分片: 2023_10_18T05_18_23.703135, latest
    • 路径: **/details_harness|drop|3_2023-10-18T05-18-23.703135.parquet
  • harness_gsm8k_5

    • 分片: 2023_10_18T05_18_23.703135, latest
    • 路径: **/details_harness|gsm8k|5_2023-10-18T05-18-23.703135.parquet
  • harness_hellaswag_10

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hellaswag|10_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: 多个路径,包括但不限于 **/details_harness|hendrycksTest-abstract_algebra|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_abstract_algebra_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-abstract_algebra|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_anatomy_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-anatomy|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_astronomy_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-astronomy|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_business_ethics_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-business_ethics|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_clinical_knowledge_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-clinical_knowledge|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_college_biology_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-college_biology|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_college_chemistry_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-college_chemistry|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_college_computer_science_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-college_computer_science|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_college_mathematics_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-college_mathematics|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_college_medicine_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-college_medicine|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_college_physics_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-college_physics|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_computer_security_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-computer_security|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_conceptual_physics_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-conceptual_physics|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_econometrics_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-econometrics|5_2023-08-30T23:37:31.114358.parquet
  • harness_hendrycksTest_electrical_engineering_5

    • 分片: 2023_08_30T23_37_31.114358, latest
    • 路径: **/details_harness|hendrycksTest-electrical_engineering|5_2023-08-30T23:37:31.114358.parquet

最新结果

以下是来自 2023-10-18T05:18:23.703135 运行的最新结果:

python { "all": { "em": 0.33598993288590606, "em_stderr": 0.004837156244460491, "f1": 0.39884333053691334, "f1_stderr": 0.00469778128078984, "acc": 0.40645791907839585, "acc_stderr": 0.00927667824750805 }, "harness|drop|3": { "em": 0.33598993288590606, "em_stderr": 0.004837156244460491, "f1": 0.39884333053691334, "f1_stderr": 0.00469778128078984 }, "harness|gsm8k|5": { "acc": 0.058377558756633814, "acc_stderr": 0.006458083557832456 }, "harness|winogrande|5": { "acc": 0.7545382794001578, "acc_stderr": 0.012095272937183644 } }

二维码
社区交流群
二维码
科研交流群
商业服务