open-llm-leaderboard/details_FlagAlpha__Llama2-Chinese-13b-Chat

该数据集是在模型FlagAlpha/Llama2-Chinese-13b-Chat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，每个配置对应一个特定的评估任务。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，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`datasets`库加载数据集的示例，并包含了特定运行的最新结果。

This dataset was automatically created during the evaluation run of the model FlagAlpha/Llama2-Chinese-13b-Chat on the Open LLM Leaderboard. The dataset consists of 64 configurations, each corresponding to a specific evaluation task. The dataset is generated from two runs, where each run is represented as a specific split for 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 used to calculate 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, and includes the most recent results from specific runs.

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 FlagAlpha/Llama2-Chinese-13b-Chat 在 Open LLM Leaderboard 上的运行过程中自动创建的。

数据集结构

• 数据集包含 64 个配置，每个配置对应一个评估任务。
• 数据集从 2 次运行中创建，每次运行可以在每个配置中找到特定的分割，分割名称使用运行的时间戳。
• "train" 分割始终指向最新的结果。
• 一个额外的配置 "results" 存储所有运行结果的聚合，用于计算和显示聚合指标在 Open LLM Leaderboard 上。

数据加载示例

python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_FlagAlpha__Llama2-Chinese-13b-Chat", "harness_winogrande_5", split="train")

最新结果

以下是 最新结果来自 run 2023-10-13T06:13:18.397506 的摘要： python { "all": { "em": 0.3886325503355705, "em_stderr": 0.004991836977358219, "f1": 0.4460098573825516, "f1_stderr": 0.004836724027731064, "acc": 0.4437472960609105, "acc_stderr": 0.010555580633054316 }, "harness|drop|3": { "em": 0.3886325503355705, "em_stderr": 0.004991836977358219, "f1": 0.4460098573825516, "f1_stderr": 0.004836724027731064 }, "harness|gsm8k|5": { "acc": 0.12585291887793784, "acc_stderr": 0.009136212598406319 }, "harness|winogrande|5": { "acc": 0.7616416732438832, "acc_stderr": 0.011974948667702313 } }

配置详情

以下是部分配置的详细信息：

• harness_arc_challenge_25

  • 分割：2023_08_17T19_12_34.146693
  • 路径：**/details_harness|arc:challenge|25_2023-08-17T19:12:34.146693.parquet
  • 分割：latest
  • 路径：**/details_harness|arc:challenge|25_2023-08-17T19:12:34.146693.parquet

• harness_drop_3

  • 分割：2023_10_13T06_13_18.397506
  • 路径：**/details_harness|drop|3_2023-10-13T06-13-18.397506.parquet
  • 分割：latest
  • 路径：**/details_harness|drop|3_2023-10-13T06-13-18.397506.parquet

• harness_gsm8k_5

  • 分割：2023_10_13T06_13_18.397506
  • 路径：**/details_harness|gsm8k|5_2023-10-13T06-13-18.397506.parquet
  • 分割：latest
  • 路径：**/details_harness|gsm8k|5_2023-10-13T06-13-18.397506.parquet

• harness_hellaswag_10

  • 分割：2023_08_17T19_12_34.146693
  • 路径：**/details_harness|hellaswag|10_2023-08-17T19:12:34.146693.parquet
  • 分割：latest
  • 路径：**/details_harness|hellaswag|10_2023-08-17T19:12:34.146693.parquet

• harness_hendrycksTest_5

  • 分割：2023_08_17T19_12_34.146693
  • 路径：**/details_harness|hendrycksTest-abstract_algebra|5_2023-08-17T19:12:34.146693.parquet 等 50 个文件
  • 分割：latest
  • 路径：**/details_harness|hendrycksTest-abstract_algebra|5_2023-08-17T19:12:34.146693.parquet 等 50 个文件