open-llm-leaderboard/details_mlabonne__ChimeraLlama-3-8B
收藏Hugging Face2024-04-23 更新2024-06-12 收录
下载链接:
https://hf-mirror.com/datasets/open-llm-leaderboard/details_mlabonne__ChimeraLlama-3-8B
下载链接
链接失效反馈官方服务:
资源简介:
该数据集是在Open LLM Leaderboard上对模型mlabonne/ChimeraLlama-3-8B进行评估时自动创建的。数据集由63个配置组成,每个配置对应一个被评估的任务。它包含一次运行的结果,每次运行在每个配置中表示为特定的分割。train分割始终指向最新结果。一个名为results的额外配置存储了所有运行的聚合结果,这些结果用于计算和显示Open LLM Leaderboard上的聚合指标。README还提供了一个示例,展示了如何使用Hugging Face的datasets库加载运行中的详细信息。
该数据集是在Open LLM Leaderboard上对模型mlabonne/ChimeraLlama-3-8B进行评估时自动创建的。数据集由63个配置组成,每个配置对应一个被评估的任务。它包含一次运行的结果,每次运行在每个配置中表示为特定的分割。train分割始终指向最新结果。一个名为results的额外配置存储了所有运行的聚合结果,这些结果用于计算和显示Open LLM Leaderboard上的聚合指标。README还提供了一个示例,展示了如何使用Hugging Face的datasets库加载运行中的详细信息。
提供机构:
open-llm-leaderboard
原始信息汇总
数据集概述
数据集摘要
数据集是在评估模型 mlabonne/ChimeraLlama-3-8B 的过程中自动创建的,用于 Open LLM Leaderboard。该数据集包含 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,每个运行可以在每个配置中找到特定的拆分,拆分名称使用运行的时间戳。"train" 拆分始终指向最新的结果。
数据集结构
数据集包含以下配置:
harness_arc_challenge_25harness_gsm8k_5harness_hellaswag_10harness_hendrycksTest_5
每个配置包含多个拆分,例如:
2024_04_23T02_44_00.437694latest
数据加载示例
python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_mlabonne__ChimeraLlama-3-8B", "harness_winogrande_5", split="train")
最新结果
最新结果来自运行 2024-04-23T02:44:00.437694,包含多个任务的评估结果,例如:
harness|arc:challenge|25harness|hellaswag|10harness|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|5harness|truthfulqa:mc|0harness|winogrande|5harness|gsm8k|5
每个任务的结果包括准确率(acc)、标准化准确率(acc_norm)、误差(acc_stderr、acc_norm_stderr)等指标。



