open-llm-leaderboard-old/details_lmsys__vicuna-13b-v1.5-16k
收藏数据集概述
数据集摘要
该数据集是在评估模型 lmsys/vicuna-13b-v1.5-16k 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。
- 数据集由 3 次运行创建,每次运行在每个配置中都有特定的分割,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_lmsys__vicuna-13b-v1.5-16k", "harness_winogrande_5", split="train")
最新结果
以下是 2023-10-15T19:40:01.175892 运行的最新结果: python { "all": { "em": 0.10056627516778524, "em_stderr": 0.003079997879762969, "f1": 0.1662111996644288, "f1_stderr": 0.003270306636878169, "acc": 0.4274584378756718, "acc_stderr": 0.010932657125643012 }, "harness|drop|3": { "em": 0.10056627516778524, "em_stderr": 0.003079997879762969, "f1": 0.1662111996644288, "f1_stderr": 0.003270306636878169 }, "harness|gsm8k|5": { "acc": 0.13115996967399546, "acc_stderr": 0.009298499235587863 }, "harness|winogrande|5": { "acc": 0.7237569060773481, "acc_stderr": 0.012566815015698158 } }
配置详情
-
harness_arc_challenge_25
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|arc:challenge|25_2023-08-09T10:54:51.508429.parquet
-
harness_drop_3
- 分割:2023_09_17T04_03_07.981708, 2023_10_15T19_40_01.175892, latest
- 路径:
**/details_harness|drop|3_2023-09-17T04-03-07.981708.parquet,**/details_harness|drop|3_2023-10-15T19-40-01.175892.parquet
-
harness_gsm8k_5
- 分割:2023_09_17T04_03_07.981708, 2023_10_15T19_40_01.175892, latest
- 路径:
**/details_harness|gsm8k|5_2023-09-17T04-03-07.981708.parquet,**/details_harness|gsm8k|5_2023-10-15T19-40-01.175892.parquet
-
harness_hellaswag_10
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hellaswag|10_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:多个路径,详见原文
-
harness_hendrycksTest_abstract_algebra_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_anatomy_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_astronomy_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_business_ethics_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_college_biology_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_college_medicine_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_college_physics_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_computer_security_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_econometrics_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-08-09T10:54:51.508429.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分割:2023_08_09T10_54_51.508429, latest
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-08-09T10:54:51.508429.parquet



