open-llm-leaderboard-old/details_Karko__Proctora
收藏数据集概述
该数据集是在模型 Karko/Proctora 在 Open LLM Leaderboard 上的评估运行期间自动创建的。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
数据集由多次运行创建,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。
此外,一个名为 "results" 的额外配置存储了所有运行的聚合结果,用于计算并在 Open LLM Leaderboard 上显示聚合指标。
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_Karko__Proctora", "harness_winogrande_5", split="train")
最新结果
以下是 2024-01-19T06:36:32.426186 运行的最新结果:
python { "all": { "acc": 0.6586351892929487, "acc_stderr": 0.03179113658240876, "acc_norm": 0.6588977596758985, "acc_norm_stderr": 0.032441095885830044, "mc1": 0.4173806609547124, "mc1_stderr": 0.017262891063272175, "mc2": 0.5955283522928707, "mc2_stderr": 0.015487512280277603 }, "harness|arc:challenge|25": { "acc": 0.6459044368600683, "acc_stderr": 0.01397545412275655, "acc_norm": 0.6783276450511946, "acc_norm_stderr": 0.013650488084494166 }, "harness|hellaswag|10": { "acc": 0.6835291774546903, "acc_stderr": 0.004641484273335098, "acc_norm": 0.866759609639514, "acc_norm_stderr": 0.0033913982936134386 }, "harness|hendrycksTest-abstract_algebra|5": { "acc": 0.35, "acc_stderr": 0.04793724854411021, "acc_norm": 0.35, "acc_norm_stderr": 0.04793724854411021 }, # 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。以下是部分配置的示例:
-
config_name: harness_arc_challenge_25
- 数据文件:
- split: 2024_01_19T06_36_32.426186
- path: **/details_harness|arc:challenge|25_2024-01-19T06-36-32.426186.parquet
- split: latest
- path: **/details_harness|arc:challenge|25_2024-01-19T06-36-32.426186.parquet
- split: 2024_01_19T06_36_32.426186
- 数据文件:
-
config_name: harness_gsm8k_5
- 数据文件:
- split: 2024_01_19T06_36_32.426186
- path: **/details_harness|gsm8k|5_2024-01-19T06-36-32.426186.parquet
- split: latest
- path: **/details_harness|gsm8k|5_2024-01-19T06-36-32.426186.parquet
- split: 2024_01_19T06_36_32.426186
- 数据文件:
-
config_name: harness_hellaswag_10
- 数据文件:
- split: 2024_01_19T06_36_32.426186
- path: **/details_harness|hellaswag|10_2024-01-19T06-36-32.426186.parquet
- split: latest
- path: **/details_harness|hellaswag|10_2024-01-19T06-36-32.426186.parquet
- split: 2024_01_19T06_36_32.426186
- 数据文件:
-
config_name: harness_hendrycksTest_5
- 数据文件:
- split: 2024_01_19T06_36_32.426186
- path:
- **/details_harness|hendrycksTest-abstract_algebra|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-anatomy|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-astronomy|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-business_ethics|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-clinical_knowledge|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-college_biology|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-college_chemistry|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-college_computer_science|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-college_mathematics|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-college_medicine|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-college_physics|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-computer_security|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-conceptual_physics|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-econometrics|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-electrical_engineering|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-elementary_mathematics|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-formal_logic|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-global_facts|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-high_school_biology|5_2024-01-19T06-36-32.426186.parquet
- **/details_harness|hendrycksTest-high_school_chemistry|5_2024-01-19T06-36-32.426186.parquet
其他数据文件...
- path:
- split: 2024_01_19T06_36_32.426186
- 数据文件:
}



