open-llm-leaderboard-old/details_rishiraj__CatPPT-base
收藏数据集概述
该数据集是在对模型 rishiraj/CatPPT-base 进行评估运行时自动创建的,用于 Open LLM Leaderboard。数据集包含 63 个配置,每个配置对应一个评估任务。
数据集结构
数据集由 1 次运行创建,每个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。此外,还有一个额外的配置 "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
以下是加载特定运行详细信息的示例代码:
python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_rishiraj__CatPPT-base", "harness_winogrande_5", split="train")
最新结果
以下是 2023-12-18T19:27:18.909562 运行的最新结果:
python { "all": { "acc": 0.6563542070521601, "acc_stderr": 0.031988233329583234, "acc_norm": 0.6566445539278223, "acc_norm_stderr": 0.03264710446236585, "mc1": 0.4357405140758874, "mc1_stderr": 0.017358345398863124, "mc2": 0.6171834778563777, "mc2_stderr": 0.015028199912315715 }, "harness|arc:challenge|25": { "acc": 0.6476109215017065, "acc_stderr": 0.013960142600598677, "acc_norm": 0.6791808873720137, "acc_norm_stderr": 0.013640943091946531 }, "harness|hellaswag|10": { "acc": 0.6739693288189603, "acc_stderr": 0.004678006403691718, "acc_norm": 0.8663612826130253, "acc_norm_stderr": 0.003395683338056335 }, "harness|hendrycksTest-abstract_algebra|5": { "acc": 0.35, "acc_stderr": 0.0479372485441102, "acc_norm": 0.35, "acc_norm_stderr": 0.0479372485441102 }, # 其他任务的结果... }
配置详情
数据集包含多个配置,每个配置对应不同的任务和数据文件路径。以下是部分配置示例:
-
harness_arc_challenge_25
- 分割:
2023_12_18T19_27_18.909562 - 路径:
**/details_harness|arc:challenge|25_2023-12-18T19-27-18.909562.parquet - 分割:
latest - 路径:
**/details_harness|arc:challenge|25_2023-12-18T19-27-18.909562.parquet
- 分割:
-
harness_gsm8k_5
- 分割:
2023_12_18T19_27_18.909562 - 路径:
**/details_harness|gsm8k|5_2023-12-18T19-27-18.909562.parquet - 分割:
latest - 路径:
**/details_harness|gsm8k|5_2023-12-18T19-27-18.909562.parquet
- 分割:
-
harness_hellaswag_10
- 分割:
2023_12_18T19_27_18.909562 - 路径:
**/details_harness|hellaswag|10_2023-12-18T19-27-18.909562.parquet - 分割:
latest - 路径:
**/details_harness|hellaswag|10_2023-12-18T19-27-18.909562.parquet
- 分割:
-
harness_hendrycksTest_5
- 分割:
2023_12_18T19_27_18.909562 - 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-anatomy|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-astronomy|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-business_ethics|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-clinical_knowledge|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-college_biology|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-college_chemistry|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-college_computer_science|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-college_mathematics|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-college_medicine|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-college_physics|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-computer_security|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-conceptual_physics|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-econometrics|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-electrical_engineering|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-elementary_mathematics|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-formal_logic|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-global_facts|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-high_school_biology|5_2023-12-18T19-27-18.909562.parquet**/details_harness|hendrycksTest-high_school_chemistry|5_2023-12-18T19-27-18.909562.parquet
其他路径...
- 分割:
}
以上是数据集的概述和部分配置详情。



