open-llm-leaderboard/details_EleutherAI__pythia-2.8b-deduped
收藏数据集概述
该数据集是在评估模型 EleutherAI/pythia-2.8b-deduped 在 Open LLM Leaderboard 上的自动创建的。数据集包含 64 个配置,每个配置对应一个评估任务。
数据集结构
- 配置数量:64 个配置
- 数据来源: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割,分割名称使用运行的时间戳。
- 最新结果:"train" 分割始终指向最新结果。
- 汇总结果: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_EleutherAI__pythia-2.8b-deduped", "harness_winogrande_5", split="train")
最新结果
以下是 2023-10-22T02:23:42.600907 运行的最新结果:
python { "all": { "em": 0.0012583892617449664, "em_stderr": 0.00036305608931190916, "f1": 0.0446549916107384, "f1_stderr": 0.0011620582208289672, "acc": 0.30527479800116447, "acc_stderr": 0.008130342870304771 }, "harness|drop|3": { "em": 0.0012583892617449664, "em_stderr": 0.00036305608931190916, "f1": 0.0446549916107384, "f1_stderr": 0.0011620582208289672 }, "harness|gsm8k|5": { "acc": 0.008339651250947688, "acc_stderr": 0.002504942226860519 }, "harness|winogrande|5": { "acc": 0.6022099447513812, "acc_stderr": 0.013755743513749023 } }
配置详情
-
harness_arc_challenge_25
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|arc:challenge|25_2023-07-19T17:26:01.712520.parquet
-
harness_drop_3
- 分割:2023_10_22T02_23_42.600907, latest
- 路径:
**/details_harness|drop|3_2023-10-22T02-23-42.600907.parquet
-
harness_gsm8k_5
- 分割:2023_10_22T02_23_42.600907, latest
- 路径:
**/details_harness|gsm8k|5_2023-10-22T02-23-42.600907.parquet
-
harness_hellaswag_10
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hellaswag|10_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:多个路径,例如
**/details_harness|hendrycksTest-abstract_algebra|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_abstract_algebra_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-abstract_algebra|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_anatomy_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-anatomy|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_astronomy_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-astronomy|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_business_ethics_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-business_ethics|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_clinical_knowledge_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-clinical_knowledge|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_college_biology_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-college_biology|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_college_chemistry_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-college_chemistry|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_college_computer_science_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-college_computer_science|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_college_mathematics_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-college_mathematics|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_college_medicine_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-college_medicine|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_college_physics_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-college_physics|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_computer_security_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-computer_security|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_conceptual_physics_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-conceptual_physics|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_econometrics_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-econometrics|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_electrical_engineering_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-electrical_engineering|5_2023-07-19T17:26:01.712520.parquet
-
harness_hendrycksTest_elementary_mathematics_5
- 分割:2023_07_19T17_26_01.712520, latest
- 路径:
**/details_harness|hendrycksTest-elementary_mathematics|5_2023-07-19T17:26:01.712520.parquet
以上是数据集的详细概述,包括数据集的结构、加载示例、最新结果以及各个配置的详细信息。



