open-llm-leaderboard/details_mosaicml__mpt-30b-instruct

该数据集是在模型 mosaicml/mpt-30b-instruct 在 Open LLM Leaderboard 上的评估运行期间自动创建的。数据集由 64 个配置组成，每个配置对应一个被评估的任务。数据集是从 2 次运行中生成的，每次运行在每个配置中表示为特定的分割。train 分割始终指向最新的结果。一个名为 results 的额外配置存储了所有运行的聚合结果，这些结果用于计算和显示 Open LLM Leaderboard 上的聚合指标。README 还提供了如何使用 Python 中的 `datasets` 库加载数据集的示例。

This dataset was automatically created during the evaluation run of the model mosaicml/mpt-30b-instruct on the Open LLM Leaderboard. It consists of 64 configurations, each corresponding to one evaluated task. The dataset is generated from two runs, with each run being represented as a specific split under each configuration. The "train" split always points to the most recent results. An additional configuration named "results" stores the aggregated results across all runs, which are utilized to compute and display the aggregate metrics on the Open LLM Leaderboard. The README also provides examples of how to load the dataset using the `datasets` library in Python.

原始信息汇总

数据集概述

数据集简介

该数据集是在评估模型 mosaicml/mpt-30b-instruct 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
• 数据分割：每个配置包含特定运行的分割，分割名称使用运行的时间戳。"train" 分割指向最新的结果。

数据加载示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_mosaicml__mpt-30b-instruct", "harness_winogrande_5", split="train")

最新结果

以下是 2023-10-14T20:57:09.846204 运行 的最新结果： python { "all": { "em": 0.3308515100671141, "em_stderr": 0.004818562129043009, "f1": 0.38283766778523554, "f1_stderr": 0.00472140525052066, "acc": 0.4522637692207808, "acc_stderr": 0.011033521433097288 }, "harness|drop|3": { "em": 0.3308515100671141, "em_stderr": 0.004818562129043009, "f1": 0.38283766778523554, "f1_stderr": 0.00472140525052066 }, "harness|gsm8k|5": { "acc": 0.15314632297194844, "acc_stderr": 0.009919728152791466 }, "harness|winogrande|5": { "acc": 0.7513812154696132, "acc_stderr": 0.012147314713403112 } }

配置详情

以下是部分配置及其数据文件路径：

• harness_arc_challenge_25

  • 分割：2023_07_20T13_11_24.937399
    • 路径：**/details_harness|arc:challenge|25_2023-07-20T13:11:24.937399.parquet
  • 分割：latest
    • 路径：**/details_harness|arc:challenge|25_2023-07-20T13:11:24.937399.parquet

• harness_drop_3

  • 分割：2023_10_14T20_57_09.846204
    • 路径：**/details_harness|drop|3_2023-10-14T20-57-09.846204.parquet
  • 分割：latest
    • 路径：**/details_harness|drop|3_2023-10-14T20-57-09.846204.parquet

• harness_gsm8k_5

  • 分割：2023_10_14T20_57_09.846204
    • 路径：**/details_harness|gsm8k|5_2023-10-14T20-57-09.846204.parquet
  • 分割：latest
    • 路径：**/details_harness|gsm8k|5_2023-10-14T20-57-09.846204.parquet

• harness_hellaswag_10

  • 分割：2023_07_20T13_11_24.937399
    • 路径：**/details_harness|hellaswag|10_2023-07-20T13:11:24.937399.parquet
  • 分割：latest
    • 路径：**/details_harness|hellaswag|10_2023-07-20T13:11:24.937399.parquet

• harness_hendrycksTest_5

  • 分割：2023_07_20T13_11_24.937399
    • 路径：
      • **/details_harness|hendrycksTest-abstract_algebra|5_2023-07-20T13:11:24.937399.parquet
      • **/details_harness|hendrycksTest-anatomy|5_2023-07-20T13:11:24.937399.parquet
      • ...（其他路径省略）
  • 分割：latest
    • 路径：
      • **/details_harness|hendrycksTest-abstract_algebra|5_2023-07-20T13:11:24.937399.parquet
      • **/details_harness|hendrycksTest-anatomy|5_2023-07-20T13:11:24.937399.parquet
      • ...（其他路径省略）