open-llm-leaderboard/details_jondurbin__airoboros-33b-gpt4-1.4
收藏数据集概述
数据集摘要
该数据集是在模型 jondurbin/airoboros-33b-gpt4-1.4 在 Open LLM Leaderboard 上的评估运行期间自动创建的。
数据集结构
- 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 3 次运行中创建,每次运行可以在每个配置中作为一个特定的分片找到,分片名称使用运行的时间戳。
- "train" 分片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 Open LLM Leaderboard 上的聚合指标。
数据加载示例
python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_jondurbin__airoboros-33b-gpt4-1.4", "harness_winogrande_5", split="train")
最新结果
以下是 2023-10-22T23:35:00.134396 运行的最新结果: python { "all": { "em": 0.008074664429530202, "em_stderr": 0.0009165188135511226, "f1": 0.07536493288590619, "f1_stderr": 0.0016308549138484363, "acc": 0.4482593962444343, "acc_stderr": 0.010265777447066696 }, "harness|drop|3": { "em": 0.008074664429530202, "em_stderr": 0.0009165188135511226, "f1": 0.07536493288590619, "f1_stderr": 0.0016308549138484363 }, "harness|gsm8k|5": { "acc": 0.11751326762699014, "acc_stderr": 0.008870331256489982 }, "harness|winogrande|5": { "acc": 0.7790055248618785, "acc_stderr": 0.011661223637643412 } }
配置详情
-
harness_arc_challenge_25
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|arc:challenge|25_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_drop_3
- 分片:
2023_10_22T20_41_55.992984,2023_10_22T23_35_00.134396,latest - 路径:
**/details_harness|drop|3_2023-10-22T20-41-55.992984.parquet,**/details_harness|drop|3_2023-10-22T23-35-00.134396.parquet
- 分片:
-
harness_gsm8k_5
- 分片:
2023_10_22T20_41_55.992984,2023_10_22T23_35_00.134396,latest - 路径:
**/details_harness|gsm8k|5_2023-10-22T20-41-55.992984.parquet,**/details_harness|gsm8k|5_2023-10-22T23-35-00.134396.parquet
- 分片:
-
harness_hellaswag_10
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hellaswag|10_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径: 多个路径,详见原文档
- 分片:
-
harness_hendrycksTest_abstract_algebra_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_anatomy_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_astronomy_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_business_ethics_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_clinical_knowledge_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_college_biology_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_college_chemistry_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_college_computer_science_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_college_mathematics_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_college_medicine_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_college_physics_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_computer_security_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_conceptual_physics_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_econometrics_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-07-31T12:50:28.372166.parquet
- 分片:
-
harness_hendrycksTest_electrical_engineering_5
- 分片:
2023_07_31T12_50_28.372166,latest - 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-07-31T12:50:28.372166.parquet
- 分片:



