open-llm-leaderboard-old/details_xxyyy123__10k_v1_lora_qkvo_rank14_v3
收藏数据集概述
数据集简介
该数据集是在评估模型 xxyyy123/10k_v1_lora_qkvo_rank14_v3 的过程中自动创建的,用于 Open LLM Leaderboard。数据集包含 61 个配置,每个配置对应一个评估任务。
数据集结构
数据集由 1 次运行创建,每个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。此外,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 Open LLM Leaderboard 上的聚合指标。
数据加载示例
以下是加载数据集详细信息的示例代码: python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_xxyyy123__10k_v1_lora_qkvo_rank14_v3", "harness_truthfulqa_mc_0", split="train")
最新结果
以下是 2023-09-03T13:17:02.987872 运行的最新结果: python { "all": { "acc": 0.5091352266849982, "acc_stderr": 0.03495474191892426, "acc_norm": 0.5128128131582483, "acc_norm_stderr": 0.03493935725866389, "mc1": 0.3671970624235006, "mc1_stderr": 0.01687480500145318, "mc2": 0.5344202990692574, "mc2_stderr": 0.015729161957393895 }, "harness|arc:challenge|25": { "acc": 0.5298634812286689, "acc_stderr": 0.014585305840007105, "acc_norm": 0.5597269624573379, "acc_norm_stderr": 0.01450676952480424 }, "harness|hellaswag|10": { "acc": 0.6050587532364071, "acc_stderr": 0.004878390226591715, "acc_norm": 0.7921728739294961, "acc_norm_stderr": 0.00404923158643323 }, "harness|hendrycksTest-abstract_algebra|5": { "acc": 0.31, "acc_stderr": 0.04648231987117316, "acc_norm": 0.31, "acc_norm_stderr": 0.04648231987117316 }, # 其他任务的结果... }
配置详情
数据集包含多个配置,每个配置对应不同的任务和数据文件路径。以下是部分配置示例:
-
配置名称:
harness_arc_challenge_25- 数据文件:
- 分割:
2023_09_03T13_17_02.987872- 路径:
**/details_harness|arc:challenge|25_2023-09-03T13:17:02.987872.parquet
- 路径:
- 分割:
latest- 路径:
**/details_harness|arc:challenge|25_2023-09-03T13:17:02.987872.parquet
- 路径:
- 分割:
- 数据文件:
-
配置名称:
harness_hellaswag_10- 数据文件:
- 分割:
2023_09_03T13_17_02.987872- 路径:
**/details_harness|hellaswag|10_2023-09-03T13:17:02.987872.parquet
- 路径:
- 分割:
latest- 路径:
**/details_harness|hellaswag|10_2023-09-03T13:17:02.987872.parquet
- 路径:
- 分割:
- 数据文件:
-
配置名称:
harness_hendrycksTest_5- 数据文件:
- 分割:
2023_09_03T13_17_02.987872- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-anatomy|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-astronomy|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-business_ethics|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-clinical_knowledge|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-college_biology|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-college_chemistry|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-college_computer_science|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-college_mathematics|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-college_medicine|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-college_physics|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-computer_security|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-conceptual_physics|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-econometrics|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-electrical_engineering|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-elementary_mathematics|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-formal_logic|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-global_facts|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-high_school_biology|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-high_school_chemistry|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-high_school_computer_science|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-high_school_european_history|5_2023-09-03T13:17:02.987872.parquet**/details_harness|hendrycksTest-high_school_geography|5_2023-09-03T13:17:02.987872.parquet- `**/details_harness|hendrycksTest-h
- 路径:
- 分割:
- 数据文件:



