five

open-llm-leaderboard/details_jondurbin__airoboros-gpt-3.5-turbo-100k-7b

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_jondurbin__airoboros-gpt-3.5-turbo-100k-7b
下载链接
链接失效反馈
官方服务:
资源简介:
该数据集是在模型评估运行期间自动创建的,包含64个配置,每个配置对应一个评估任务。数据集由多个运行生成,每个运行作为一个特定分割,以运行的时间戳命名。train分割始终指向最新结果。另外,results配置存储了用于计算和在Open LLM排行榜上显示的聚合结果。

该数据集是在模型评估运行期间自动创建的,包含64个配置,每个配置对应一个评估任务。数据集由多个运行生成,每个运行作为一个特定分割,以运行的时间戳命名。train分割始终指向最新结果。另外,results配置存储了用于计算和在Open LLM排行榜上显示的聚合结果。
提供机构:
open-llm-leaderboard
原始信息汇总

数据集概述

数据集是在评估模型 jondurbin/airoboros-gpt-3.5-turbo-100k-7bOpen LLM Leaderboard 上的运行过程中自动创建的。

数据集组成

  • 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。每次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。
  • 一个额外的配置 "results" 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。

加载数据集示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_jondurbin__airoboros-gpt-3.5-turbo-100k-7b", "harness_winogrande_5", split="train")

最新结果

以下是 2023-10-18T13:27:38.857487 运行的最新结果: python { "all": { "em": 0.04236577181208054, "em_stderr": 0.0020627521235254487, "f1": 0.12625943791946267, "f1_stderr": 0.002580178700995439, "acc": 0.3718023208847917, "acc_stderr": 0.008942653172749105 }, "harness|drop|3": { "em": 0.04236577181208054, "em_stderr": 0.0020627521235254487, "f1": 0.12625943791946267, "f1_stderr": 0.002580178700995439 }, "harness|gsm8k|5": { "acc": 0.0356330553449583, "acc_stderr": 0.005106107853744191 }, "harness|winogrande|5": { "acc": 0.7079715864246251, "acc_stderr": 0.012779198491754018 } }

配置详情

  • harness_arc_challenge_25

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|arc:challenge|25_2023-08-03T10:51:43.635262.parquet
  • harness_drop_3

    • 分片:2023_10_18T13_27_38.857487, latest
    • 路径:**/details_harness|drop|3_2023-10-18T13-27-38.857487.parquet
  • harness_gsm8k_5

    • 分片:2023_10_18T13_27_38.857487, latest
    • 路径:**/details_harness|gsm8k|5_2023-10-18T13-27-38.857487.parquet
  • harness_hellaswag_10

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hellaswag|10_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:多个路径,例如 **/details_harness|hendrycksTest-abstract_algebra|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_abstract_algebra_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-abstract_algebra|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_anatomy_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-anatomy|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_astronomy_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-astronomy|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_business_ethics_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-business_ethics|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_clinical_knowledge_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-clinical_knowledge|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_college_biology_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-college_biology|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_college_chemistry_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-college_chemistry|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_college_computer_science_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-college_computer_science|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_college_mathematics_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-college_mathematics|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_college_medicine_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-college_medicine|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_college_physics_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-college_physics|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_computer_security_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-computer_security|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_conceptual_physics_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-conceptual_physics|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_econometrics_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-econometrics|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_electrical_engineering_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-electrical_engineering|5_2023-08-03T10:51:43.635262.parquet
  • harness_hendrycksTest_elementary_mathematics_5

    • 分片:2023_08_03T10_51_43.635262, latest
    • 路径:**/details_harness|hendrycksTest-elementary_mathematics|5_2023-08-03T10:51:43.635262.parquet

以上是数据集的详细概述,包括数据集的组成、加载示例、最新结果以及各个配置的详细信息。

二维码
社区交流群
二维码
科研交流群
商业服务