open-llm-leaderboard-old/details_WangZeJun__bloom-820m-chat
收藏数据集概述
数据集简介
该数据集是在对模型 WangZeJun/bloom-820m-chat 进行评估运行期间自动创建的,用于 Open LLM Leaderboard。
数据集结构
- 配置数量:64个配置,每个配置对应一个评估任务。
- 创建来源:数据集从2次运行中创建,每次运行在每个配置中作为一个特定的分割,分割名称使用运行的时间戳。
- 最新结果:"train" 分割总是指向最新的结果。
- 结果汇总:一个额外的配置 "results" 存储所有运行的汇总结果,用于计算和显示在 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_WangZeJun__bloom-820m-chat", "harness_winogrande_5", split="train")
最新结果
以下是 2023-09-17T22:00:08.030398 运行的最新结果: python { "all": { "em": 0.0382760067114094, "em_stderr": 0.0019648445106113157, "f1": 0.08853187919463057, "f1_stderr": 0.0023716202448817885, "acc": 0.265982636148382, "acc_stderr": 0.007011869610583192 }, "harness|drop|3": { "em": 0.0382760067114094, "em_stderr": 0.0019648445106113157, "f1": 0.08853187919463057, "f1_stderr": 0.0023716202448817885 }, "harness|gsm8k|5": { "acc": 0.0, "acc_stderr": 0.0 }, "harness|winogrande|5": { "acc": 0.531965272296764, "acc_stderr": 0.014023739221166384 } }
配置详情
-
harness_arc_challenge_25
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|arc:challenge|25_2023-08-17T10:54:24.303970.parquet
-
harness_drop_3
- 分割:2023_09_17T22_00_08.030398, latest
- 路径:
**/details_harness|drop|3_2023-09-17T22-00-08.030398.parquet
-
harness_gsm8k_5
- 分割:2023_09_17T22_00_08.030398, latest
- 路径:
**/details_harness|gsm8k|5_2023-09-17T22-00-08.030398.parquet
-
harness_hellaswag_10
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hellaswag|10_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:多个文件路径,例如
**/details_harness|hendrycksTest-abstract_algebra|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_abstract_algebra_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_anatomy_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_astronomy_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_business_ethics_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_college_biology_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_college_medicine_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_college_physics_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_computer_security_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_econometrics_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-08-17T10:54:24.303970.parquet
-
harness_hendrycksTest_elementary_mathematics_5
- 分割:2023_08_17T10_54_24.303970, latest
- 路径:
**/details_harness|hendrycksTest-elementary_mathematics|5_2023-08-17T10:54:24.303970.parquet



