open-llm-leaderboard/details_codellama__CodeLlama-7b-hf
收藏数据集概述
数据集来源
该数据集是在对模型 codellama/CodeLlama-7b-hf 进行评估运行期间自动创建的,评估结果发布在 Open LLM Leaderboard 上。
数据集结构
- 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。
- 额外的 "results" 配置存储所有运行的聚合结果,用于计算和显示聚合指标在 Open LLM Leaderboard 上。
数据加载示例
python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_codellama__CodeLlama-7b-hf", "harness_winogrande_5", split="train")
最新结果
以下是 2023-10-14T19:46:33.225068 运行的最新结果: python { "all": { "em": 0.0006291946308724832, "em_stderr": 0.00025680027497238217, "f1": 0.05123741610738289, "f1_stderr": 0.001242998424746743, "acc": 0.34582445982552373, "acc_stderr": 0.009790248772764803 }, "harness|drop|3": { "em": 0.0006291946308724832, "em_stderr": 0.00025680027497238217, "f1": 0.05123741610738289, "f1_stderr": 0.001242998424746743 }, "harness|gsm8k|5": { "acc": 0.05155420773313116, "acc_stderr": 0.006090887955262816 }, "harness|winogrande|5": { "acc": 0.6400947119179163, "acc_stderr": 0.01348960959026679 } }
配置详情
-
harness_arc_challenge_25
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|arc:challenge|25_2023-08-26T04:20:17.128606.parquet
-
harness_drop_3
- 分割: 2023_10_14T19_46_33.225068, latest
- 路径: **/details_harness|drop|3_2023-10-14T19-46-33.225068.parquet
-
harness_gsm8k_5
- 分割: 2023_10_14T19_46_33.225068, latest
- 路径: **/details_harness|gsm8k|5_2023-10-14T19-46-33.225068.parquet
-
harness_hellaswag_10
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hellaswag|10_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: 包含多个子任务的 parquet 文件
-
harness_hendrycksTest_abstract_algebra_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-abstract_algebra|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_anatomy_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-anatomy|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_astronomy_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-astronomy|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_business_ethics_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-business_ethics|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-clinical_knowledge|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_college_biology_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-college_biology|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-college_chemistry|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-college_computer_science|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-college_mathematics|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_college_medicine_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-college_medicine|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_college_physics_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-college_physics|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_computer_security_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-computer_security|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-conceptual_physics|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_econometrics_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-econometrics|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-electrical_engineering|5_2023-08-26T04:20:17.128606.parquet
-
harness_hendrycksTest_elementary_mathematics_5
- 分割: 2023_08_26T04_20_17.128606, latest
- 路径: **/details_harness|hendrycksTest-elementary_mathematics|5_2023-08-26T04:20:17.128606.parquet
以上是数据集的详细概述,包括数据集的来源、结构、加载示例、最新结果以及各个配置的详细信息。



