open-llm-leaderboard-old/details_ewqr2130__llama2-ppo
收藏数据集概述
数据集摘要
该数据集是在对模型 ewqr2130/llama2-ppo 进行评估运行期间自动创建的,用于 Open LLM Leaderboard。数据集包含 63 个配置,每个配置对应一个评估任务。数据集从 1 次运行中创建,每个运行的详细信息可以在每个配置中找到,使用运行的时间戳作为分割名称。"train" 分割始终指向最新的结果。
数据集结构
数据集包含多个配置,每个配置对应不同的评估任务。以下是部分配置的示例:
-
harness_arc_challenge_25
- 分割:2024_01_05T00_23_47.259679
- 路径:
**/details_harness|arc:challenge|25_2024-01-05T00-23-47.259679.parquet - 分割:latest
- 路径:
**/details_harness|arc:challenge|25_2024-01-05T00-23-47.259679.parquet
-
harness_gsm8k_5
- 分割:2024_01_05T00_23_47.259679
- 路径:
**/details_harness|gsm8k|5_2024-01-05T00-23-47.259679.parquet - 分割:latest
- 路径:
**/details_harness|gsm8k|5_2024-01-05T00-23-47.259679.parquet
-
harness_hellaswag_10
- 分割:2024_01_05T00_23_47.259679
- 路径:
**/details_harness|hellaswag|10_2024-01-05T00-23-47.259679.parquet - 分割:latest
- 路径:
**/details_harness|hellaswag|10_2024-01-05T00-23-47.259679.parquet
-
harness_hendrycksTest_5
- 分割:2024_01_05T00_23_47.259679
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-anatomy|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-astronomy|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-business_ethics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-clinical_knowledge|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-college_biology|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-college_chemistry|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-college_computer_science|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-college_mathematics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-college_medicine|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-college_physics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-computer_security|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-conceptual_physics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-econometrics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-electrical_engineering|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-elementary_mathematics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-formal_logic|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-global_facts|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_biology|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_chemistry|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_computer_science|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_european_history|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_geography|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_government_and_politics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_macroeconomics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_mathematics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_microeconomics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_physics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_psychology|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_statistics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_us_history|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-high_school_world_history|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-human_aging|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-human_sexuality|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-international_law|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-jurisprudence|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-logical_fallacies|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-machine_learning|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-management|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-marketing|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-medical_genetics|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-miscellaneous|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-moral_disputes|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-moral_scenarios|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-nutrition|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-philosophy|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-prehistory|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-professional_accounting|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-professional_law|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-professional_medicine|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-professional_psychology|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-public_relations|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-security_studies|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-sociology|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-us_foreign_policy|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-virology|5_2024-01-05T00-23-47.259679.parquet**/details_harness|hendrycksTest-world_religions|5_2024-01-05T00-23-47.259679.parquet
最新结果
以下是 2024-01-05T00:23:47.259679 运行 的最新结果:
python { "all": { "acc": 0.3526851673994734, "acc_stderr": 0.03310876929515637, "acc_norm": 0.35709972795834366, "acc_norm_stderr": 0.03399609567460545, "mc1": 0.22643818849449204, "mc1_stderr": 0.014651337324602597, "mc2": 0.4507763893909204, "mc2_stderr": 0.016309761592194282 }, "harness|arc:challenge|25": { "acc": 0.36177474402730375, "acc_stderr": 0.014041957945038071, "acc_norm": 0.4



