lesserfield/lmsys-arena-human-preference-winner-43k-unfiltered

---
license: apache-2.0
task_categories:
- text-generation
language:
- en
pretty_name: LMSYS Chatbot Arena Human Preference Winner Unfiltered
size_categories:
- 10K<n<100K
---

# lmsys-arena-human-preference-winner-43k-unfiltered

This repository contains a dataset derived from the [lmsys/lmsys-arena-human-preference-55k](https://huggingface.co/datasets/lmsys/lmsys-arena-human-preference-55k) dataset, which is licensed under the Apache 2.0 License.

## Dataset Description

The `lmsys-arena-human-preference-winner-43k-unfiltered` dataset is a collection of 43,000 samples, each containing an instruction (prompt) and an output (winning response) from real-world user and LLM conversations. The dataset is derived from the original `lmsys-arena-human-preference-55k` dataset, with the following modifications:

1. **Data Structure Change**: The original `train.csv` file has been converted to a `train_clean.jsonl` file, where each line represents a sample in the format `{ "instruction": prompt, "output": response }`.
2. **Winner Selection**: For each sample, the winning response (as determined by the user preference label in the original dataset) has been selected and assigned to the `output` field.
3. **Unfiltering**: The dataset has been unfiltered by removing unwanted words.
```python
unwanted_words = [
"text-based ai language model",
"domestic violence",
"please refrain",
"derogatory",
"inappropriate",
"offensive",
"racism",
"racist",
"racial",
"discriminate",
"discriminatory",
"discrimination",
"sexist",
"sexism",
"unacceptable",
"inclusive workplace",
"lgbt",
"morals",
"ethics",
"ethical",
"legality",
"illegal",
"illegality",
"hateful",
"harmful",
"it is never okay",
"it is important to",
"it's important to",
"real-world consequences",
"hate speech",
"glorify",
"not be appropriate",
"supremacist",
"extremist",
"responsible ai",
"ai principles",
"ai assistant",
"an ai language",
"ableist",
"hurtful",
"gender stereotype",
"gender inequality",
"underrepresentation",
"safe spaces",
"gender-based",
"inclusivity",
"feminist",
"feminism",
"transgender",
"empowerment",
"communist",
"capitalism",
"stereotypes",
"biases",
"bias",
"microaggression",
"prioritize human safety",
"as a language model",
"as an ai language model",
"as a large language model",
"as an ai",
"ethical principles",
"consensual",
"it is not appropriate",
"it's not appropriate",
"i cannot fulfill your request",
"harmful to human beings",
"ethical guidelines",
"my guidelines",
"prioritize user safety",
"adhere to ethical guidelines",
"harmful consequences",
"potentially harmful",
"dangerous activities",
"promote safety",
"well-being of all users",
"responsible information sharing",
"jeopardize the safety",
"illegal actions or intentions",
"undermine the stability",
"promote the well-being",
"illegal activities or actions",
"adherence to the law",
"potentially be harmful",
"illegal substances or activities",
"committed to promoting",
"safe information",
"lawful information",
"cannot provide guidance",
"cannot provide information",
"unable to offer assistance",
"cannot engage in discussions",
"programming prohibits",
"follow ethical guidelines",
"ensure the safety",
"involves an illegal subject",
"prioritize safety",
"illegal subject",
"prioritize user well-being",
"cannot support or promote",
"activities that could harm",
"pose a risk to others",
"against my programming",
"activities that could undermine",
"potentially dangerous",
"not within the scope",
"designed to prioritize safety",
"not able to provide",
"maintain user safety",
"adhere to safety guidelines",
"dangerous or harmful",
"cannot provide any information",
"focus on promoting safety",
"an ai language model you don't have",
"as an ai language model, i cannot",
"as an ai language model, i do not",
"as an ai language model, i am not able",
"as an ai language model, i don't have personal",
"i am an ai language model and do not",
"however, it is important to use any code or information provided responsibly and within legal and ethical boundaries.",
"as an ai language model, i don't have",
"as an ai language model, i am only able",
"ai language model and i do not",
"as an ai language model, i cannot modify",
"as an ai language model, i do not",
"i know as an ai language model you don't have",
"as an ai language model, you cannot",
"i'm sorry, but as an ai language model",
"as an ai language model, i don't have",
"unfortunately, i cannot provide",
"i'm sorry, i cannot",
"i'm sorry, i cannot generate",
"ai cannot create or program",
"i'm afraid i cannot create",
"you cannot create an",
"it operates ethically and is",
"had an ethical system",
"ensuring the ethical",
"and ethical sourcing",
"are from ethical",
"legal and ethical",
"engage in unethical",
"unethical or aggressive",
"unethical business",
"como modelo de lenguaje ai",
"lo siento, como modelo de lenguaje",
"no puedo proporcionar",
"pero debido a mi capacidad para generar c\u00f3digos complejos y completos es limitado",
"lo siento, pero no puedo",
"lo siento, pero como modelo de lenguaje, no puedo proporcionar",
"lo siento, como modelo de lenguaje, no tengo",
"lo siento, debe haber habido una confusi\u00f3n",
"lo siento, como modelo de lenguaje, no puedo realizar",
"lo siento, soy un modelo de lenguaje y no tengo la capacidad de generar",
"lamento no poder proporcionarte el c\u00f3digo",
"desculpe-me, mas a linguagem vulgar e ofensiva",
"apropriada em nenhum contexto",
"como modelo de linguagem",
"como um modelo de linguagem, n\u00e3o tenho a capacidade de",
"i cannot assist",
"prioritize ethical",
"respectful",
"morally",
"i'm sorry,",
"i'm an",
"i am an",
"i'm an ai" ,
"i am an ai",
"my purpose",
"filter_bad_language",
"filter\_bad\_language",
"entertainment purposes",
"purely hypothetical",
"not a human",
"i am an ai",
"cannot provide",
"can't provide",
"won't provide",
"not provide",
"worth noting",
"cause harm",
"a language model",
"keep in mind",
"unethical",
"bad language",
"the words ****",
"bad_language",
"certainly not",
"complying",
"comply",
"i cannot",
"my main goal",
"as a machine",
"i don't have the ability",
"i am here to assist",
"my purpose is to ",
"my knowledge cutoff",
"my knowledge cut off",
"september 2021",
"regulations",
"not be suitable",
"i apologize, but",
"it is not possible",
"controversial",
"my programming",
"ethically",
"it is important to",
"please note",
"sensitive topic",
"not acceptable",
"it is important for",
"divisive",
"not appropriate",
"our values",
"f\*cking",
"f\*ck",
"sh\*t",
"diversity and",
"diversity and inclusion",
"values diversity",
"social responsibility",
"environmental, social, and governance",
" esg ",
"against women",
"problematic history",
"diversity",
"*this chat conversation is shared from",
"*this conversation is shared from",
"i can't assist",
"as an ai language model, i don't",
"against the terms of service",
"i'm sorry, but"
]
```
## License

This dataset is derived from the `lmsys/lmsys-arena-human-preference-55k` dataset, which is licensed under the Apache 2.0 License. As such, this derived dataset is also licensed under the Apache 2.0 License.

## Disclaimer

This dataset is provided "as is" without any warranty or guarantees. The dataset may contain biases, inaccuracies, or inappropriate content. It is the responsibility of the user to review and evaluate the suitability of the dataset for their intended use case.

## Citation

If you use this dataset in your research or project, please cite the original dataset:

```bibtex
@dataset{lmsys_arena_human_preference_55k,
author = {LMSys},
title = {lmsys-arena-human-preference-55k},
url = {https://huggingface.co/datasets/lmsys/lmsys-arena-human-preference-55k},
license = {Apache 2.0},
year = {2024}
}
```

## Acknowledgments

This derived dataset is based on the work of LMSys and the contributors to the original `lmsys-arena-human-preference-55k` dataset. We express our gratitude for their efforts in creating and sharing this valuable resource.

The `lmsys-arena-human-preference-winner-43k-unfiltered` dataset is derived from the `lmsys/lmsys-arena-human-preference-55k` dataset and contains 43,000 samples. Each sample includes an instruction (prompt) and an output (winning response) from real-world user and LLM conversations. The dataset has undergone modifications such as data structure changes, winner selection based on user preference, and unfiltering by removing unwanted words. The dataset is licensed under the Apache 2.0 License.

原始信息汇总

lmsys-arena-human-preference-winner-43k-unfiltered

数据集描述

lmsys-arena-human-preference-winner-43k-unfiltered 数据集包含 43,000 个样本，每个样本包含一个指令（提示）和一个输出（获胜响应），来自真实用户和 LLM 对话。该数据集源自 lmsys-arena-human-preference-55k 数据集，并进行了以下修改：

1. 数据结构变化：原始的 train.csv 文件已转换为 train_clean.jsonl 文件，每行代表一个样本，格式为 { "instruction": prompt, "output": response }。
2. 获胜响应选择：每个样本中，根据原始数据集中的用户偏好标签选择获胜响应，并分配给 output 字段。
3. 去过滤化：数据集已去过滤化，移除了不需要的词汇。

许可证

该数据集基于 lmsys/lmsys-arena-human-preference-55k 数据集，该数据集在 Apache 2.0 许可证下发布。因此，该衍生数据集也在 Apache 2.0 许可证下发布。

免责声明

该数据集按“原样”提供，不提供任何保证或担保。数据集可能包含偏见、不准确或不适当的内容。用户有责任审查和评估数据集对其预期用例的适用性。

引用

如果您在研究或项目中使用此数据集，请引用原始数据集：

bibtex @dataset{lmsys_arena_human_preference_55k, author = {LMSys}, title = {lmsys-arena-human-preference-55k}, url = {https://huggingface.co/datasets/lmsys/lmsys-arena-human-preference-55k}, license = {Apache 2.0}, year = {2024} }

致谢

该衍生数据集基于 LMSys 和原始 lmsys-arena-human-preference-55k 数据集的贡献者的工作。我们对此表示感谢。