open-llm-leaderboard/details_sia-ai__llama-2-7b-1-percent-open-orca-1000-steps-v0

该数据集是在模型sia-ai/llama-2-7b-1-percent-open-orca-1000-steps-v0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，每个配置对应一个评估任务。它包含2次运行的数据，每次运行在每个配置中表示为特定的分割。train分割始终指向最新结果。一个名为results的额外配置存储了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。README还提供了如何使用Hugging Face datasets库加载数据集的示例。This dataset was automatically created during the evaluation run of the model sia-ai/llama-2-7b-1-percent-open-orca-1000-steps-v0 on the Open LLM Leaderboard. The dataset consists of 64 configurations, each corresponding to one evaluation task. It contains data from 2 runs, with each run represented as a specific split within each configuration. The train split always points to the most recent results. An additional configuration named results stores the aggregated results from all runs, which are used to calculate and display the aggregate metrics on the Open LLM Leaderboard. The README also provides examples of how to load the dataset using the Hugging Face datasets library.

原始信息汇总

数据集概述

该数据集是在对模型 sia-ai/llama-2-7b-1-percent-open-orca-1000-steps-v0 进行评估运行期间自动创建的，用于 Open LLM Leaderboard 的评估。

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。
• 数据集由 2 次运行创建，每次运行可以在每个配置中找到特定的分割，分割名称使用运行的时间戳。
• "train" 分割始终指向最新的结果。
• 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。

数据加载示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_sia-ai__llama-2-7b-1-percent-open-orca-1000-steps-v0", "harness_winogrande_5", split="train")

最新结果

以下是来自 2023-10-23T07:22:00.656974 运行的最新结果（注意可能包含其他任务的结果，每个任务的结果可以在 "results" 和 "latest" 分割中找到）：

python { "all": { "em": 0.08672399328859061, "em_stderr": 0.0028821083852410504, "f1": 0.13875838926174472, "f1_stderr": 0.0030398596200052856, "acc": 0.384207080894677, "acc_stderr": 0.00839930559560157 }, "harness|drop|3": { "em": 0.08672399328859061, "em_stderr": 0.0028821083852410504, "f1": 0.13875838926174472, "f1_stderr": 0.0030398596200052856 }, "harness|gsm8k|5": { "acc": 0.027293404094010616, "acc_stderr": 0.004488095380209766 }, "harness|winogrande|5": { "acc": 0.7411207576953434, "acc_stderr": 0.012310515810993376 } }

配置详情

• harness_arc_challenge_25

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|arc:challenge|25_2023-08-18T13:55:57.511446.parquet

• harness_drop_3

  • 分割: 2023_10_23T07_22_00.656974, latest
  • 路径: **/details_harness|drop|3_2023-10-23T07-22-00.656974.parquet

• harness_gsm8k_5

  • 分割: 2023_10_23T07_22_00.656974, latest
  • 路径: **/details_harness|gsm8k|5_2023-10-23T07-22-00.656974.parquet

• harness_hellaswag_10

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hellaswag|10_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: 多个路径，例如 **/details_harness|hendrycksTest-abstract_algebra|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_abstract_algebra_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-abstract_algebra|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_anatomy_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-anatomy|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_astronomy_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-astronomy|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_business_ethics_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-business_ethics|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_clinical_knowledge_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-clinical_knowledge|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_college_biology_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-college_biology|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_college_chemistry_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-college_chemistry|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_college_computer_science_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-college_computer_science|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_college_mathematics_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-college_mathematics|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_college_medicine_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-college_medicine|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_college_physics_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-college_physics|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_computer_security_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-computer_security|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_conceptual_physics_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-conceptual_physics|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_econometrics_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-econometrics|5_2023-08-18T13:55:57.511446.parquet

• harness_hendrycksTest_electrical_engineering_5

  • 分割: 2023_08_18T13_55_57.511446, latest
  • 路径: **/details_harness|hendrycksTest-electrical_engineering|5_2023-08-18T13:55:57.511446.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。