open-llm-leaderboard/details_HWERI__pythia-70m-deduped-cleansharegpt
收藏数据集概述
该数据集是在模型 HWERI/pythia-70m-deduped-cleansharegpt 的评估运行期间自动创建的。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。
数据集结构
数据集包含以下配置:
-
harness_arc_challenge_25
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|arc:challenge|25_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|arc:challenge|25_2023-09-13T12-28-53.949092.parquet
-
harness_drop_3
- 分割:2023_10_29T05_46_53.221413
- 路径:
**/details_harness|drop|3_2023-10-29T05-46-53.221413.parquet - 分割:latest
- 路径:
**/details_harness|drop|3_2023-10-29T05-46-53.221413.parquet
-
harness_gsm8k_5
- 分割:2023_10_29T05_46_53.221413
- 路径:
**/details_harness|gsm8k|5_2023-10-29T05-46-53.221413.parquet - 分割:latest
- 路径:
**/details_harness|gsm8k|5_2023-10-29T05-46-53.221413.parquet
-
harness_hellaswag_10
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hellaswag|10_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hellaswag|10_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-13T12-28-53.949092.parquet等 41 个文件 - 分割:latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-13T12-28-53.949092.parquet等 41 个文件
-
harness_hendrycksTest_abstract_algebra_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_anatomy_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_astronomy_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_business_ethics_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_college_biology_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_college_medicine_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_college_physics_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_computer_security_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_econometrics_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-09-13T12-28-53.949092.parquet
-
harness_hendrycksTest_elementary_mathematics_5
- 分割:2023_09_13T12_28_53.949092
- 路径:
**/details_harness|hendrycksTest-elementary_mathematics|5_2023-09-13T12-28-53.949092.parquet - 分割:latest
- 路径:
**/details_harness|hendrycksTest-elementary_mathematics|5_2023-09-13T12-28-53.949092.parquet
最新结果
以下是 2023-10-29T05:46:53.221413 运行 的最新结果:
python { "all": { "em": 0.0, "em_stderr": 0.0, "f1": 0.0, "f1_stderr": 0.0, "acc": 0.2600631412786109, "acc_stderr": 0.007020548332172166 }, "harness|drop|3": { "em": 0.0, "em_stderr": 0.0, "f1": 0.0, "f1_stderr": 0.0 }, "harness|gsm8k|5": { "acc": 0.0, "acc_stderr": 0.0 }, "harness|winogrande|5": { "acc": 0.5201262825572218, "acc_stderr": 0.014041096664344332 } }



