open-llm-leaderboard-old/details_PY007__TinyLlama-1.1B-step-50K-105b

该数据集是在Open LLM Leaderboard上对模型PY007/TinyLlama-1.1B-step-50K-105b进行评估时自动生成的。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次运行生成，每次运行在每个配置中表示为特定的分割，分割名称由运行的时间戳命名。train分割始终指向最新的结果。此外，还有一个results配置存储了所有运行的聚合结果，用于在Open LLM Leaderboard上计算和显示聚合指标。README还提供了如何使用Hugging Face datasets库加载数据集的示例。

原始信息汇总

数据集概述

数据集名称

• Evaluation run of PY007/TinyLlama-1.1B-step-50K-105b

数据集来源

• 该数据集是在模型 PY007/TinyLlama-1.1B-step-50K-105b 在 Open LLM Leaderboard 上的评估运行期间自动创建的。

数据集组成

• 该数据集包含 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，用于计算并在 Open LLM Leaderboard 上显示聚合指标。

数据加载示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_PY007__TinyLlama-1.1B-step-50K-105b", "harness_winogrande_5", split="train")

最新结果

• 这些是最新结果来自 run 2023-10-24T19:38:12.477684： python { "all": { "em": 0.001153523489932886, "em_stderr": 0.0003476179896857113, "f1": 0.039109689597315506, "f1_stderr": 0.001116278403063508, "acc": 0.27455565641618196, "acc_stderr": 0.0079998796659362 }, "harness|drop|3": { "em": 0.001153523489932886, "em_stderr": 0.0003476179896857113, "f1": 0.039109689597315506, "f1_stderr": 0.001116278403063508 }, "harness|gsm8k|5": { "acc": 0.00530705079605762, "acc_stderr": 0.0020013057209480774 }, "harness|winogrande|5": { "acc": 0.5438042620363063, "acc_stderr": 0.013998453610924324 } }

配置详情

• harness_arc_challenge_25

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|arc:challenge|25_2023-09-12T12-30-04.204611.parquet

• harness_drop_3

  • 分割：2023_10_24T19_38_12.477684, latest
  • 路径：**/details_harness|drop|3_2023-10-24T19-38-12.477684.parquet

• harness_gsm8k_5

  • 分割：2023_10_24T19_38_12.477684, latest
  • 路径：**/details_harness|gsm8k|5_2023-10-24T19-38-12.477684.parquet

• harness_hellaswag_10

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hellaswag|10_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：多个路径，例如 **/details_harness|hendrycksTest-abstract_algebra|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_abstract_algebra_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_anatomy_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-anatomy|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_astronomy_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-astronomy|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_business_ethics_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-business_ethics|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_clinical_knowledge_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-clinical_knowledge|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_college_biology_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-college_biology|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_college_chemistry_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-college_chemistry|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_college_computer_science_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-college_computer_science|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_college_mathematics_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-college_mathematics|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_college_medicine_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-college_medicine|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_college_physics_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-college_physics|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_computer_security_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-computer_security|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_conceptual_physics_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-conceptual_physics|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_econometrics_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-econometrics|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_electrical_engineering_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-electrical_engineering|5_2023-09-12T12-30-04.204611.parquet

• harness_hendrycksTest_elementary_mathematics_5

  • 分割：2023_09_12T12_30_04.204611, latest
  • 路径：**/details_harness|hendrycksTest-elementary_mathematics|5_2023-09-12T12-30-04.204611.parquet