open-llm-leaderboard-old/details_lu-vae__llama2-13B-sharegpt4-orca-openplatypus-8w
收藏数据集概述
该数据集是在模型lu-vae/llama2-13B-sharegpt4-orca-openplatypus-8w的评估运行期间自动创建的,用于Open LLM Leaderboard。
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_lu-vae__llama2-13B-sharegpt4-orca-openplatypus-8w", "harness_winogrande_5", split="train")
最新结果
以下是2023-10-25T02:53:14.013945运行的最新结果:
python { "all": { "em": 0.002936241610738255, "em_stderr": 0.0005541113054709996, "f1": 0.0881931627516778, "f1_stderr": 0.0017824166498732977, "acc": 0.43444724154830167, "acc_stderr": 0.01050882298494655 }, "harness|drop|3": { "em": 0.002936241610738255, "em_stderr": 0.0005541113054709996, "f1": 0.0881931627516778, "f1_stderr": 0.0017824166498732977 }, "harness|gsm8k|5": { "acc": 0.11751326762699014, "acc_stderr": 0.008870331256489993 }, "harness|winogrande|5": { "acc": 0.7513812154696132, "acc_stderr": 0.012147314713403108 } }
配置详情
-
harness_arc_challenge_25
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|arc:challenge|25_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_drop_3
- 分割:
2023_10_25T02_53_14.013945,latest - 路径:
**/details_harness|drop|3_2023-10-25T02-53-14.013945.parquet
- 分割:
-
harness_gsm8k_5
- 分割:
2023_10_25T02_53_14.013945,latest - 路径:
**/details_harness|gsm8k|5_2023-10-25T02-53-14.013945.parquet
- 分割:
-
harness_hellaswag_10
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hellaswag|10_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径: 多个路径,例如
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_abstract_algebra_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_anatomy_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_astronomy_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_business_ethics_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_clinical_knowledge_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_college_biology_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_college_chemistry_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_college_computer_science_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_college_mathematics_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_college_medicine_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_college_physics_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_computer_security_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_conceptual_physics_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_econometrics_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-09-14T08-19-15.271974.parquet
- 分割:
-
harness_hendrycksTest_electrical_engineering_5
- 分割:
2023_09_14T08_19_15.271974,latest - 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-09-14T08-19-15.271974.parquet
- 分割:
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。



