open-llm-leaderboard/details_winglian__basilisk-4b
收藏数据集概述
该数据集是在对模型 winglian/basilisk-4b 进行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_winglian__basilisk-4b", "harness_winogrande_5", split="train")
最新结果
以下是 2023-10-24T05:37:13.853267 运行的最新结果: python { "all": { "em": 0.005453020134228188, "em_stderr": 0.000754172779679255, "f1": 0.03888842281879202, "f1_stderr": 0.001263593140660679, "acc": 0.26558800315706393, "acc_stderr": 0.007012571320319758 }, "harness|drop|3": { "em": 0.005453020134228188, "em_stderr": 0.000754172779679255, "f1": 0.03888842281879202, "f1_stderr": 0.001263593140660679 }, "harness|gsm8k|5": { "acc": 0.0, "acc_stderr": 0.0 }, "harness|winogrande|5": { "acc": 0.5311760063141279, "acc_stderr": 0.014025142640639516 } }
配置详情
配置列表
harness_arc_challenge_25harness_drop_3harness_gsm8k_5harness_hellaswag_10harness_hendrycksTest_5harness_hendrycksTest_abstract_algebra_5harness_hendrycksTest_anatomy_5harness_hendrycksTest_astronomy_5harness_hendrycksTest_business_ethics_5harness_hendrycksTest_clinical_knowledge_5harness_hendrycksTest_college_biology_5harness_hendrycksTest_college_chemistry_5harness_hendrycksTest_college_computer_science_5harness_hendrycksTest_college_mathematics_5harness_hendrycksTest_college_medicine_5harness_hendrycksTest_college_physics_5harness_hendrycksTest_computer_security_5harness_hendrycksTest_conceptual_physics_5harness_hendrycksTest_econometrics_5harness_hendrycksTest_electrical_engineering_5harness_hendrycksTest_elementary_mathematics_5
数据文件路径
每个配置包含多个数据文件,路径和分割如下:
harness_arc_challenge_25- 分割:
2023_10_03T14_16_48.676759- 路径:
**/details_harness|arc:challenge|25_2023-10-03T14-16-48.676759.parquet
- 路径:
- 分割:
latest- 路径:
**/details_harness|arc:challenge|25_2023-10-03T14-16-48.676759.parquet
- 路径:
- 分割:
harness_drop_3- 分割:
2023_10_24T05_37_13.853267- 路径:
**/details_harness|drop|3_2023-10-24T05-37-13.853267.parquet
- 路径:
- 分割:
latest- 路径:
**/details_harness|drop|3_2023-10-24T05-37-13.853267.parquet
- 路径:
- 分割:
harness_gsm8k_5- 分割:
2023_10_24T05_37_13.853267- 路径:
**/details_harness|gsm8k|5_2023-10-24T05-37-13.853267.parquet
- 路径:
- 分割:
latest- 路径:
**/details_harness|gsm8k|5_2023-10-24T05-37-13.853267.parquet
- 路径:
- 分割:
harness_hellaswag_10- 分割:
2023_10_03T14_16_48.676759- 路径:
**/details_harness|hellaswag|10_2023-10-03T14-16-48.676759.parquet
- 路径:
- 分割:
latest- 路径:
**/details_harness|hellaswag|10_2023-10-03T14-16-48.676759.parquet
- 路径:
- 分割:
harness_hendrycksTest_5- 分割:
2023_10_03T14_16_48.676759- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-elementary_mathematics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-formal_logic|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-global_facts|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_biology|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_chemistry|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_computer_science|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_european_history|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_geography|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_government_and_politics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_macroeconomics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_mathematics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_microeconomics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_physics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_psychology|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_statistics|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_us_history|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-high_school_world_history|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-human_aging|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-human_sexuality|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-international_law|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-jurisprudence|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-logical_fallacies|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-machine_learning|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-management|5_2023-10-03T14-16-48.676759.parquet - 路径:
**/details_harness|hendrycksTest-marketing|5_2023-10-03T14-16-48.676759.parquet - 路径:`**/details_harness|hendrycksTest-medical_genetics|5_2023-10-03T
- 路径:
- 分割:



