open-llm-leaderboard/details_Undi95__Unholy-v1-12L-13B
收藏数据集概述
数据集来源
该数据集是在评估模型 Undi95/Unholy-v1-12L-13B 在 Open LLM Leaderboard 上的运行过程中自动创建的。
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。
- 运行次数:数据集由2次运行创建。每次运行在每个配置中作为一个特定的分片存在,分片名称使用运行的时间戳。
- 最新结果:"train" 分片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_Undi95__Unholy-v1-12L-13B", "harness_winogrande_5", split="train")
最新结果
以下是 2023-10-29T08:07:07.360378 运行 的最新结果: python { "all": { "em": 0.022651006711409395, "em_stderr": 0.0015237307803438198, "f1": 0.09728712248322129, "f1_stderr": 0.00210132435826052, "acc": 0.44169065680213837, "acc_stderr": 0.010210392359241776 }, "harness|drop|3": { "em": 0.022651006711409395, "em_stderr": 0.0015237307803438198, "f1": 0.09728712248322129, "f1_stderr": 0.00210132435826052 }, "harness|gsm8k|5": { "acc": 0.1106899166034875, "acc_stderr": 0.008642172551392465 }, "harness|winogrande|5": { "acc": 0.7726913970007893, "acc_stderr": 0.011778612167091087 } }
配置详情
-
harness_arc_challenge_25
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|arc:challenge|25_2023-09-18T13-52-19.375562.parquet
-
harness_drop_3
- 分片:2023_10_29T08_07_07.360378, latest
- 路径:
**/details_harness|drop|3_2023-10-29T08-07-07.360378.parquet
-
harness_gsm8k_5
- 分片:2023_10_29T08_07_07.360378, latest
- 路径:
**/details_harness|gsm8k|5_2023-10-29T08-07-07.360378.parquet
-
harness_hellaswag_10
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hellaswag|10_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:多个路径,例如
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_abstract_algebra_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_anatomy_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_astronomy_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_business_ethics_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_college_biology_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_college_medicine_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_college_physics_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_computer_security_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_econometrics_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-09-18T13-52-19.375562.parquet
-
harness_hendrycksTest_elementary_mathematics_5
- 分片:2023_09_18T13_52_19.375562, latest
- 路径:
**/details_harness|hendrycksTest-elementary_mathematics|5_2023-09-18T13-52-19.375562.parquet
以上是数据集的详细概述,包括数据集的来源、结构、加载示例、最新结果以及各个配置的详细信息。



