open-llm-leaderboard-old/details_Aspik101__StableBeluga-13B-instruct-PL-lora_unload
收藏数据集概述
该数据集是在评估模型 Aspik101/StableBeluga-13B-instruct-PL-lora_unload 在 Open LLM Leaderboard 上的自动创建的。
数据集组成
- 数据集包含 64 个配置,每个配置对应一个评估任务。
- 数据集从 2 次运行中创建,每次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" 分割总是指向最新的结果。
- 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。
最新结果
- 最新结果来自 2023-10-13T13:27:06.134462 运行,包含以下指标: python { "all": { "em": 0.014786073825503355, "em_stderr": 0.0012360366760472946, "f1": 0.08531564597315425, "f1_stderr": 0.001909056545524939, "acc": 0.44382508573319457, "acc_stderr": 0.010461592536002241 }, "harness|drop|3": { "em": 0.014786073825503355, "em_stderr": 0.0012360366760472946, "f1": 0.08531564597315425, "f1_stderr": 0.001909056545524939 }, "harness|gsm8k|5": { "acc": 0.12206216830932524, "acc_stderr": 0.009017054965766495 }, "harness|winogrande|5": { "acc": 0.7655880031570639, "acc_stderr": 0.011906130106237986 } }
配置详情
-
harness_arc_challenge_25
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|arc:challenge|25_2023-08-09T11:43:44.316126.parquet
-
harness_drop_3
- 分割: 2023_10_13T13_27_06.134462, latest
- 路径:
**/details_harness|drop|3_2023-10-13T13-27-06.134462.parquet
-
harness_gsm8k_5
- 分割: 2023_10_13T13_27_06.134462, latest
- 路径:
**/details_harness|gsm8k|5_2023-10-13T13-27-06.134462.parquet
-
harness_hellaswag_10
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hellaswag|10_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径: 多个路径,例如
**/details_harness|hendrycksTest-abstract_algebra|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_abstract_algebra_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_anatomy_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_astronomy_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_business_ethics_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_college_biology_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_college_medicine_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_college_physics_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_computer_security_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_econometrics_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-08-09T11:43:44.316126.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分割: 2023_08_09T11_43_44.316126, latest
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-08-09T11:43:44.316126.parquet



