open-llm-leaderboard-old/details_TheBloke__orca_mini_13B-GPTQ

---
pretty_name: Evaluation run of TheBloke/orca_mini_13B-GPTQ
dataset_summary: "Dataset automatically created during the evaluation run of model\
\ [TheBloke/orca_mini_13B-GPTQ](https://huggingface.co/TheBloke/orca_mini_13B-GPTQ)\
\ on the [Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard).\n\
\nThe dataset is composed of 3 configuration, each one coresponding to one of the\
\ evaluated task.\n\nThe dataset has been created from 2 run(s). Each run can be\
\ found as a specific split in each configuration, the split being named using the\
\ timestamp of the run.The \"train\" split is always pointing to the latest results.\n\
\nAn additional configuration \"results\" store all the aggregated results of the\
\ run (and is used to compute and display the aggregated metrics on the [Open LLM\
\ Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)).\n\
\nTo load the details from a run, you can for instance do the following:\n```python\n\
from datasets import load_dataset\ndata = load_dataset(\"open-llm-leaderboard/details_TheBloke__orca_mini_13B-GPTQ_public\"\
,\n\t\"harness_winogrande_5\",\n\tsplit=\"train\")\n```\n\n## Latest results\n\n\
These are the [latest results from run 2023-11-07T10:33:18.298818](https://huggingface.co/datasets/open-llm-leaderboard/details_TheBloke__orca_mini_13B-GPTQ_public/blob/main/results_2023-11-07T10-33-18.298818.json)(note\
\ that their might be results for other tasks in the repos if successive evals didn't\
\ cover the same tasks. You find each in the results and the \"latest\" split for\
\ each eval):\n\n```python\n{\n \"all\": {\n \"em\": 0.04047818791946309,\n\
\ \"em_stderr\": 0.002018262301743542,\n \"f1\": 0.11770658557046992,\n\
\ \"f1_stderr\": 0.002544480345951201,\n \"acc\": 0.3192425320418652,\n\
\ \"acc_stderr\": 0.007133502794987517\n },\n \"harness|drop|3\": {\n\
\ \"em\": 0.04047818791946309,\n \"em_stderr\": 0.002018262301743542,\n\
\ \"f1\": 0.11770658557046992,\n \"f1_stderr\": 0.002544480345951201\n\
\ },\n \"harness|gsm8k|5\": {\n \"acc\": 0.000758150113722517,\n \
\ \"acc_stderr\": 0.0007581501137225239\n },\n \"harness|winogrande|5\"\
: {\n \"acc\": 0.6377269139700079,\n \"acc_stderr\": 0.01350885547625251\n\
\ }\n}\n```"
repo_url: https://huggingface.co/TheBloke/orca_mini_13B-GPTQ
leaderboard_url: https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard
point_of_contact: [e-mail]
configs:
- config_name: harness_drop_3
data_files:
- split: 2023_11_05T13_43_32.201116
path:
- '**/details_harness|drop|3_2023-11-05T13-43-32.201116.parquet'
- split: 2023_11_07T10_33_18.298818
path:
- '**/details_harness|drop|3_2023-11-07T10-33-18.298818.parquet'
- split: latest
path:
- '**/details_harness|drop|3_2023-11-07T10-33-18.298818.parquet'
- config_name: harness_gsm8k_5
data_files:
- split: 2023_11_05T13_43_32.201116
path:
- '**/details_harness|gsm8k|5_2023-11-05T13-43-32.201116.parquet'
- split: 2023_11_07T10_33_18.298818
path:
- '**/details_harness|gsm8k|5_2023-11-07T10-33-18.298818.parquet'
- split: latest
path:
- '**/details_harness|gsm8k|5_2023-11-07T10-33-18.298818.parquet'
- config_name: harness_winogrande_5
data_files:
- split: 2023_11_05T13_43_32.201116
path:
- '**/details_harness|winogrande|5_2023-11-05T13-43-32.201116.parquet'
- split: 2023_11_07T10_33_18.298818
path:
- '**/details_harness|winogrande|5_2023-11-07T10-33-18.298818.parquet'
- split: latest
path:
- '**/details_harness|winogrande|5_2023-11-07T10-33-18.298818.parquet'
- config_name: results
data_files:
- split: 2023_11_05T13_43_32.201116
path:
- results_2023-11-05T13-43-32.201116.parquet
- split: 2023_11_07T10_33_18.298818
path:
- results_2023-11-07T10-33-18.298818.parquet
- split: latest
path:
- results_2023-11-07T10-33-18.298818.parquet
---

# Dataset Card for Evaluation run of TheBloke/orca_mini_13B-GPTQ

## Dataset Description

- **Homepage:**
- **Repository:** https://huggingface.co/TheBloke/orca_mini_13B-GPTQ
- **Paper:**
- **Leaderboard:** https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard
- **Point of Contact:** [e-mail]

### Dataset Summary

Dataset automatically created during the evaluation run of model [TheBloke/orca_mini_13B-GPTQ](https://huggingface.co/TheBloke/orca_mini_13B-GPTQ) on the [Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard).

The dataset is composed of 3 configuration, each one coresponding to one of the evaluated task.

The dataset has been created from 2 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the latest results.

An additional configuration "results" store all the aggregated results of the run (and is used to compute and display the aggregated metrics on the [Open LLM Leaderboard](https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard)).

To load the details from a run, you can for instance do the following:
```python
from datasets import load_dataset
data = load_dataset("open-llm-leaderboard/details_TheBloke__orca_mini_13B-GPTQ_public",
"harness_winogrande_5",
split="train")
```

## Latest results

These are the [latest results from run 2023-11-07T10:33:18.298818](https://huggingface.co/datasets/open-llm-leaderboard/details_TheBloke__orca_mini_13B-GPTQ_public/blob/main/results_2023-11-07T10-33-18.298818.json)(note that their might be results for other tasks in the repos if successive evals didn't cover the same tasks. You find each in the results and the "latest" split for each eval):

```python
{
"all": {
"em": 0.04047818791946309,
"em_stderr": 0.002018262301743542,
"f1": 0.11770658557046992,
"f1_stderr": 0.002544480345951201,
"acc": 0.3192425320418652,
"acc_stderr": 0.007133502794987517
},
"harness|drop|3": {
"em": 0.04047818791946309,
"em_stderr": 0.002018262301743542,
"f1": 0.11770658557046992,
"f1_stderr": 0.002544480345951201
},
"harness|gsm8k|5": {
"acc": 0.000758150113722517,
"acc_stderr": 0.0007581501137225239
},
"harness|winogrande|5": {
"acc": 0.6377269139700079,
"acc_stderr": 0.01350885547625251
}
}
```

### Supported Tasks and Leaderboards

[More Information Needed]

### Languages

[More Information Needed]

## Dataset Structure

### Data Instances

[More Information Needed]

### Data Fields

[More Information Needed]

### Data Splits

[More Information Needed]

## Dataset Creation

### Curation Rationale

[More Information Needed]

### Source Data

#### Initial Data Collection and Normalization

[More Information Needed]

#### Who are the source language producers?

[More Information Needed]

### Annotations

#### Annotation process

[More Information Needed]

#### Who are the annotators?

[More Information Needed]

### Personal and Sensitive Information

[More Information Needed]

## Considerations for Using the Data

### Social Impact of Dataset

[More Information Needed]

### Discussion of Biases

[More Information Needed]

### Other Known Limitations

[More Information Needed]

## Additional Information

### Dataset Curators

[More Information Needed]

### Licensing Information

[More Information Needed]

### Citation Information

[More Information Needed]

### Contributions

[More Information Needed]

原始信息汇总

数据集卡片 for Evaluation run of TheBloke/orca_mini_13B-GPTQ

数据集描述

数据集摘要

数据集是在模型 TheBloke/orca_mini_13B-GPTQ 在 Open LLM Leaderboard 上的评估运行期间自动创建的。

数据集由 3 个配置组成，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。

额外的配置 "results" 存储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）。

加载运行细节的示例如下： python from datasets import load_dataset data = load_dataset("open-llm-leaderboard/details_TheBloke__orca_mini_13B-GPTQ_public", "harness_winogrande_5", split="train")

最新结果

以下是 2023-11-07T10:33:18.298818 运行的最新结果（注意，如果连续评估未覆盖相同任务，仓库中可能会有其他任务的结果。您可以在 "results" 和每个评估的 "latest" 分割中找到每个任务的结果）：

python { "all": { "em": 0.04047818791946309, "em_stderr": 0.002018262301743542, "f1": 0.11770658557046992, "f1_stderr": 0.002544480345951201, "acc": 0.3192425320418652, "acc_stderr": 0.007133502794987517 }, "harness|drop|3": { "em": 0.04047818791946309, "em_stderr": 0.002018262301743542, "f1": 0.11770658557046992, "f1_stderr": 0.002544480345951201 }, "harness|gsm8k|5": { "acc": 0.000758150113722517, "acc_stderr": 0.0007581501137225239 }, "harness|winogrande|5": { "acc": 0.6377269139700079, "acc_stderr": 0.01350885547625251 } }

数据集结构

配置

• harness_drop_3

  • 分割: 2023_11_05T13_43_32.201116
    • 路径: **/details_harness|drop|3_2023-11-05T13-43-32.201116.parquet
  • 分割: 2023_11_07T10_33_18.298818
    • 路径: **/details_harness|drop|3_2023-11-07T10-33-18.298818.parquet
  • 分割: latest
    • 路径: **/details_harness|drop|3_2023-11-07T10-33-18.298818.parquet

• harness_gsm8k_5

  • 分割: 2023_11_05T13_43_32.201116
    • 路径: **/details_harness|gsm8k|5_2023-11-05T13-43-32.201116.parquet
  • 分割: 2023_11_07T10_33_18.298818
    • 路径: **/details_harness|gsm8k|5_2023-11-07T10-33-18.298818.parquet
  • 分割: latest
    • 路径: **/details_harness|gsm8k|5_2023-11-07T10-33-18.298818.parquet

• harness_winogrande_5

  • 分割: 2023_11_05T13_43_32.201116
    • 路径: **/details_harness|winogrande|5_2023-11-05T13-43-32.201116.parquet
  • 分割: 2023_11_07T10_33_18.298818
    • 路径: **/details_harness|winogrande|5_2023-11-07T10-33-18.298818.parquet
  • 分割: latest
    • 路径: **/details_harness|winogrande|5_2023-11-07T10-33-18.298818.parquet

• results

  • 分割: 2023_11_05T13_43_32.201116
    • 路径: results_2023-11-05T13-43-32.201116.parquet
  • 分割: 2023_11_07T10_33_18.298818
    • 路径: results_2023-11-07T10-33-18.298818.parquet
  • 分割: latest
    • 路径: results_2023-11-07T10-33-18.298818.parquet

搜集汇总

数据集介绍

构建方式

在大型语言模型评估领域，数据集构建的自动化与标准化是确保结果可比性的关键。本数据集是在Open LLM Leaderboard平台上，对模型TheBloke/orca_mini_13B-GPTQ进行自动化评估运行时动态生成的。其构建过程完全依托于标准化的评估流程，将每一次评估运行的结果自动捕获并结构化存储。数据集由三个核心配置构成，分别对应DROP、GSM8K和Winogrande三项评测任务，每个配置内又包含以时间戳命名的具体运行切分，确保了评估历史的完整追溯。此外，一个独立的“results”配置专门用于汇总并存储所有运行的聚合指标，为模型性能的宏观分析提供了数据基础。

特点

作为模型评估过程的直接产物，该数据集的核心特征体现在其高度的结构化和时序性上。数据集并非静态集合，而是动态记录了模型在特定时间点于多项基准任务上的详细表现，每一次评估运行都作为一个独立的切分被保存，使得研究者能够纵向对比模型在不同迭代阶段的性能演变。其结构清晰地区分了任务配置与结果汇总，便于进行细粒度的任务分析或整体性能评估。数据集中包含的精确率、F1分数等指标及其标准误，为量化模型能力与不确定性提供了可靠依据，体现了评估工作的严谨性与透明度。

使用方法

对于希望深入分析模型评估细节的研究者而言，该数据集提供了便捷的访问接口。通过Hugging Face的`datasets`库，用户可以灵活加载特定任务配置下的数据。例如，调用`load_dataset`函数并指定数据集名称、配置（如`harness_winogrande_5`）以及切分（如`latest`或具体时间戳），即可获取相应的评估细节数据。对于聚合结果，可直接访问“results”配置。这种设计使得用户既能聚焦于单一任务的详细输出进行微观考察，也能便捷地获取模型的综合性能指标，服务于模型比较、性能诊断或评估方法学研究等多种场景。

背景与挑战

背景概述

在大型语言模型（LLM）迅猛发展的时代背景下，评估模型的综合能力成为推动技术演进的关键环节。HuggingFace社区于2023年推出的Open LLM Leaderboard，旨在为各类开源语言模型提供一个标准化、透明的性能评测平台。该数据集作为该排行榜的衍生产物，专门记录了模型TheBloke/orca_mini_13B-GPTQ在DROP、GSM8K和Winogrande等多项核心评测任务中的详细表现数据。其创建源于对模型量化后性能影响的深入探究，通过自动化流程捕获了不同时间点的评估结果，为研究社区提供了模型能力演进的微观视角，助力于模型优化与比较研究的深入开展。

当前挑战

该数据集所应对的核心挑战在于如何精准评估经过量化压缩的大型语言模型在复杂推理与常识理解任务上的性能保持度。量化技术虽能提升模型部署效率，但往往伴随性能损失，尤其在数学推理（如GSM8K）和阅读理解（如DROP）等需要精确数值处理的任务中，模型表现可能显著下降，这构成了模型优化领域的关键难题。在数据集构建层面，挑战体现在确保评估过程的可复现性与结果的一致性。由于评测依赖于动态的自动化流程，如何在不同运行间维持评测环境的稳定，并妥善处理版本迭代可能带来的任务覆盖差异，是保证数据可比性与权威性的技术难点。

常用场景

经典使用场景

在大语言模型评估领域，该数据集作为Open LLM Leaderboard评估流程的产物，其经典使用场景在于为研究者提供orca_mini_13B-GPTQ模型在多个标准基准任务上的详细性能数据。通过整合DROP、GSM8K和Winogrande等评测配置，该数据集能够系统性地展示模型在阅读理解、数学推理和常识推理等核心自然语言处理任务上的表现，为模型间的横向对比与性能剖析提供了结构化依据。

解决学术问题

该数据集直接回应了开放领域大语言模型标准化评估的学术需求。它通过记录模型在多项任务上的精确指标，如准确率、F1分数及其标准误，为解决模型能力量化、性能可复现性以及评估流程透明化等关键研究问题提供了数据支撑。其意义在于构建了一个可追溯的评估档案，促进了学术社区对模型优劣的客观评判，并推动了评估方法论向更严谨、更细粒度的方向发展。

衍生相关工作

围绕此类源自Open LLM Leaderboard的评估数据集，已衍生出多项经典研究工作。例如，基于多任务评估结果进行的模型能力溯源分析，探究模型在不同任务上表现差异的内在机理；利用时序评估数据开展的模型性能演进研究，刻画模型随着训练或微调而产生的动态变化；此外，这些详尽的评估数据也常被用作元分析或综述研究的基础，用于总结特定模型架构或训练范式的整体效能与局限。