open-llm-leaderboard/details_TehVenom__GPT-J-Pyg_PPO-6B-Dev-V8p4
收藏数据集概述
数据集来源
该数据集是在对模型 TehVenom/GPT-J-Pyg_PPO-6B-Dev-V8p4 进行评估运行期间自动创建的,评估运行在 Open LLM Leaderboard 上进行。
数据集组成
数据集由64个配置组成,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。
结果汇总
一个额外的配置 "results" 存储所有运行结果的汇总(用于计算和显示 Open LLM Leaderboard 上的汇总指标)。
数据加载示例
以下是加载特定运行详细信息的示例代码: python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_TehVenom__GPT-J-Pyg_PPO-6B-Dev-V8p4", "harness_winogrande_5", split="train")
最新结果
以下是 2023-10-18T06:55:06.084057 运行 的最新结果: python { "all": { "em": 0.0016778523489932886, "em_stderr": 0.0004191330178826871, "f1": 0.058036912751678014, "f1_stderr": 0.001339441597906354, "acc": 0.3295242323804896, "acc_stderr": 0.008622843965649133 }, "harness|drop|3": { "em": 0.0016778523489932886, "em_stderr": 0.0004191330178826871, "f1": 0.058036912751678014, "f1_stderr": 0.001339441597906354 }, "harness|gsm8k|5": { "acc": 0.018953752843062926, "acc_stderr": 0.0037560783410314704 }, "harness|winogrande|5": { "acc": 0.6400947119179163, "acc_stderr": 0.013489609590266795 } }
数据集配置
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_5harness_hendrycksTest_formal_logic_5harness_hendrycksTest_global_facts_5harness_hendrycksTest_high_school_biology_5harness_hendrycksTest_high_school_chemistry_5harness_hendrycksTest_high_school_computer_science_5harness_hendrycksTest_high_school_european_history_5harness_hendrycksTest_high_school_geography_5harness_hendrycksTest_high_school_government_and_politics_5harness_hendrycksTest_high_school_macroeconomics_5harness_hendrycksTest_high_school_mathematics_5harness_hendrycksTest_high_school_microeconomics_5harness_hendrycksTest_high_school_physics_5harness_hendrycksTest_high_school_psychology_5harness_hendrycksTest_high_school_statistics_5harness_hendrycksTest_high_school_us_history_5harness_hendrycksTest_high_school_world_history_5harness_hendrycksTest_human_aging_5harness_hendrycksTest_human_sexuality_5harness_hendrycksTest_international_law_5harness_hendrycksTest_jurisprudence_5harness_hendrycksTest_logical_fallacies_5harness_hendrycksTest_machine_learning_5harness_hendrycksTest_management_5harness_hendrycksTest_marketing_5harness_hendrycksTest_medical_genetics_5harness_hendrycksTest_miscellaneous_5harness_hendrycksTest_moral_disputes_5harness_hendrycksTest_moral_scenarios_5harness_hendrycksTest_nutrition_5harness_hendrycksTest_philosophy_5harness_hendrycksTest_prehistory_5harness_hendrycksTest_professional_accounting_5harness_hendrycksTest_professional_law_5harness_hendrycksTest_professional_medicine_5harness_hendrycksTest_professional_psychology_5harness_hendrycksTest_public_relations_5harness_hendrycksTest_security_studies_5harness_hendrycksTest_sociology_5harness_hendrycksTest_us_foreign_policy_5harness_hendrycksTest_virology_5harness_hendrycksTest_world_religions_5
数据文件路径
每个配置包含多个数据文件路径,具体路径信息请参考原始文档。



