open-llm-leaderboard/details_Delcos__Mistral-Pygmalion-7b
收藏数据集概述
数据集简介
该数据集是在对模型 Delcos/Mistral-Pygmalion-7b 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。
- 运行次数:数据集从2次运行中创建。每个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。
- 训练分片:"train" 分片总是指向最新的结果。
- 结果配置: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_Delcos__Mistral-Pygmalion-7b", "harness_winogrande_5", split="train")
最新结果
这些是最新的结果,来自2023年10月28日13:05:25.339926的运行: python { "all": { "em": 0.001363255033557047, "em_stderr": 0.0003778609196460774, "f1": 0.05936241610738259, "f1_stderr": 0.0013656193493625718, "acc": 0.41059662883495607, "acc_stderr": 0.009533380943461503 }, "harness|drop|3": { "em": 0.001363255033557047, "em_stderr": 0.0003778609196460774, "f1": 0.05936241610738259, "f1_stderr": 0.0013656193493625718 }, "harness|gsm8k|5": { "acc": 0.06823351023502654, "acc_stderr": 0.006945358944067431 }, "harness|winogrande|5": { "acc": 0.7529597474348856, "acc_stderr": 0.012121402942855575 } }
配置详情
以下是部分配置的详细信息:
-
harness_arc_challenge_25
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|arc:challenge|25_2023-10-10T20-14-17.715432.parquet
-
harness_drop_3
- 分片:2023_10_28T13_05_25.339926, latest
- 路径:
**/details_harness|drop|3_2023-10-28T13-05-25.339926.parquet
-
harness_gsm8k_5
- 分片:2023_10_28T13_05_25.339926, latest
- 路径:
**/details_harness|gsm8k|5_2023-10-28T13-05-25.339926.parquet
-
harness_hellaswag_10
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hellaswag|10_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:多个路径,例如
**/details_harness|hendrycksTest-abstract_algebra|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_abstract_algebra_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_anatomy_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_astronomy_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_business_ethics_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_college_biology_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_college_medicine_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_college_physics_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_computer_security_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_econometrics_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-10-10T20-14-17.715432.parquet
-
harness_hendrycksTest_elementary_mathematics_5
- 分片:2023_10_10T20_14_17.715432, latest
- 路径:
**/details_harness|hendrycksTest-elementary_mathematics|5_2023-10-10T20-14-17.715432.parquet



