open-llm-leaderboard-old/details_nomic-ai__gpt4all-j
收藏数据集概述
数据集简介
该数据集是在评估模型 nomic-ai/gpt4all-j 在 Open LLM Leaderboard 上的运行过程中自动创建的。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
数据集由 3 次运行结果组成,每次运行的详细结果可以在每个配置中找到,以运行的时间戳命名的特定分割形式存储。"train" 分割始终指向最新的结果。
额外配置
一个额外的配置 "results" 存储了所有运行的聚合结果,用于计算并在 Open LLM Leaderboard 上显示聚合指标。
数据加载示例
以下是一个加载特定运行详细结果的示例代码: python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_nomic-ai__gpt4all-j", "harness_truthfulqa_mc_0", split="train")
最新结果
以下是 2023-10-03T19:47:17.747407 运行 的最新结果示例: python { "all": { "acc": 0.2872188662344035, "acc_stderr": 0.03256180276717163, "acc_norm": 0.2903884169662422, "acc_norm_stderr": 0.03256130411949783, "mc1": 0.2827417380660955, "mc1_stderr": 0.015764770836777305, "mc2": 0.4277581711709451, "mc2_stderr": 0.014665895347989117 }, "harness|arc:challenge|25": { "acc": 0.38993174061433444, "acc_stderr": 0.014252959848892884, "acc_norm": 0.4197952218430034, "acc_norm_stderr": 0.014422181226303022 }, "harness|hellaswag|10": { "acc": 0.483469428400717, "acc_stderr": 0.004987053652540272, "acc_norm": 0.6406094403505278, "acc_norm_stderr": 0.004788412062375707 }, "harness|hendrycksTest-abstract_algebra|5": { "acc": 0.27, "acc_stderr": 0.0446196043338474, "acc_norm": 0.27, "acc_norm_stderr": 0.0446196043338474 }, # 其他任务的结果... }
配置详情
数据集包含多个配置,每个配置对应不同的任务和运行结果。以下是部分配置的示例:
-
config_name: harness_arc_challenge_25
- split: 2023_08_23T12_03_22.271414
- path: **/details_harness|arc:challenge|25_2023-08-23T12:03:22.271414.parquet
- split: 2023_10_03T19_47_17.747407
- path: **/details_harness|arc:challenge|25_2023-10-03T19-47-17.747407.parquet
- split: latest
- path: **/details_harness|arc:challenge|25_2023-10-03T19-47-17.747407.parquet
- split: 2023_08_23T12_03_22.271414
-
config_name: harness_drop_3
- split: 2023_09_18T00_56_07.678298
- path: **/details_harness|drop|3_2023-09-18T00-56-07.678298.parquet
- split: latest
- path: **/details_harness|drop|3_2023-09-18T00-56-07.678298.parquet
- split: 2023_09_18T00_56_07.678298
-
config_name: harness_gsm8k_5
- split: 2023_09_18T00_56_07.678298
- path: **/details_harness|gsm8k|5_2023-09-18T00-56-07.678298.parquet
- split: latest
- path: **/details_harness|gsm8k|5_2023-09-18T00-56-07.678298.parquet
- split: 2023_09_18T00_56_07.678298
-
config_name: harness_hellaswag_10
- split: 2023_08_23T12_03_22.271414
- path: **/details_harness|hellaswag|10_2023-08-23T12:03:22.271414.parquet
- split: 2023_10_03T19_47_17.747407
- path: **/details_harness|hellaswag|10_2023-10-03T19-47-17.747407.parquet
- split: latest
- path: **/details_harness|hellaswag|10_2023-10-03T19-47-17.747407.parquet
- split: 2023_08_23T12_03_22.271414
-
config_name: harness_hendrycksTest_5
- split: 2023_08_23T12_03_22.271414
- path:
- **/details_harness|hendrycksTest-abstract_algebra|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-anatomy|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-astronomy|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-business_ethics|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-clinical_knowledge|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-college_biology|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-college_chemistry|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-college_computer_science|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-college_mathematics|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-college_medicine|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-college_physics|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-computer_security|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-conceptual_physics|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-econometrics|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-electrical_engineering|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-elementary_mathematics|5_2023-08-23T12:03:22.271414.parquet
- **/details_harness|hendrycksTest-formal_logic|5_2023-08
- path:
- split: 2023_08_23T12_03_22.271414



