PersonaHub Keyword Annotations

This dataset contains the first 25 000 personas from the proj-persona/PersonaHub elite_persona config. Each persona has been tagged with keywords generated using the agentlans/flan-t5-small-keywords model. This dataset could be useful for looking up and generating personas related to a given topic.

  Limitations

The original dataset contains personas generated en masse and some may be inconsistent or of uneven quality The keyword… See the full description on the dataset page: https://huggingface.co/datasets/agentlans/PersonaHub-keywords.

Dataset

This dataset was created by ATHviii

Contents

saumyamalik/personahub-code-v2-34999-unused-gemma3 dataset hosted on Hugging Face and contributed by the HF Datasets community

saumyamalik/personahub-code-v2-34999-unused-qwq dataset hosted on Hugging Face and contributed by the HF Datasets community

Dataset Card for PersonaHub FineWeb-Edu 4 Embeddings

This dataset has been created with distilabel.

  Dataset Summary

This dataset obtains embeddings for the dataset argilla-warehouse/personahub-fineweb-edu-4-dedup, using the Alibaba-NLP/gte-large-en-v1.5 model from sentence transformers. The pipeline can be seen at: pipe_personahub_embeddings.py.

  Dataset structure

This dataset contains a pipeline.yaml which can be used to reproduce the pipeline… See the full description on the dataset page: https://huggingface.co/datasets/argilla-warehouse/personahub-fineweb-edu-4-embeddings.

Dataset Card for personahub-fineweb-edu-comparison

This dataset has been created with distilabel. The pipeline script was uploaded to easily reproduce the dataset: pipe_personahub_compare.py. It can be run directly using the CLI: distilabel pipeline run --script "https://huggingface.co/datasets/plaguss/personahub-fineweb-edu-comparison/raw/main/pipe_personahub_compare.py"

  Dataset Summary

This dataset contains a pipeline.yaml which can be used to… See the full description on the dataset page: https://huggingface.co/datasets/plaguss/personahub-fineweb-edu-comparison.

GPT-4 Generated Data Reference:

(Original) https://huggingface.co/datasets/proj-persona/PersonaHub (Original Github) https://github.com/tencent-ailab/persona-hub

PersonaHub with Business Keywords

Filter from https://huggingface.co/datasets/proj-persona/PersonaHub I am using this code to filter the dataset from datasets import load_dataset, Dataset import random

Load the original dataset

dataset = load_dataset('proj-persona/PersonaHub','persona',split='train')

def filter_keywords(example): keywords=["business","management"] persona = example['persona'].lower() return any(keyword in persona for keyword in keywords)

Apply the… See the full description on the dataset page: https://huggingface.co/datasets/pacozaa/PersonaHub_business.

Translated proj-persona/PersonaHub using nayohan/llama3-instrucTrans-enko-8b. For this dataset, we only used data that is 5000 characters or less in length and has language of English. Thanks for @proj-persona and @nayohan.

  Scaling Synthetic Data Creation with 1,000,000,000 Personas

This repo releases data introduced in our paper Scaling Synthetic Data Creation with 1,000,000,000 Personas: We propose a novel persona-driven data synthesis methodology that leverages various… See the full description on the dataset page: https://huggingface.co/datasets/youjunhyeok/PersonaHub-ko.

This repository contains the datasets and materials used to analyze and replicate the results presented in our paper investigating how persona-based prompting affects the political orientations of Large Language Models (LLMs).

Contents

The repository includes files organized by model (Mistral, Llama, Qwen, and Zephyr) and experimental condition (base, right-authoritarian [ra], and left-libertarian [ll]):

Model Response Data

• *_persona_compass_base.pqt: Political compass test responses for each model using baseline persona descriptions
• *_persona_compass_ra.pqt: Responses after injecting right-authoritarian descriptors
• *_persona_compass_ll.pqt: Responses after injecting left-libertarian descriptors

Configuration and Input Files

• personas.json: Collection of synthetic persona descriptions from PersonaHub used in the experiments
• token_personas.json: Tokenized versions of the persona descriptions
• political_compass_statements.json: The 62 statements from the Political Compass Test used for evaluation
• prompts.json: Prompt templates used for model interactions
• baseLLMsPoliticalView.json: Default political orientations of the models without persona prompting

Related Code Repository

The code used to analyze this data and reproduce the results presented in the paper can be found at: https://github.com/d-lab/llm-political-personas

File Placement Instructions

After downloading, organize the files as follows:

Configuration and Input Files

Place all the configuration files in the data/raw/ directory.

Model Response Files

Rename all model-specific .pqt files to persona_compass.pqt and place them in their respective directories:

• Base condition files:
  • data/processed/Llama-3.1-8B-Instruct/base/persona_compass.pqt
  • data/processed/Mistral-7B-Instruct-v0.3/base/persona_compass.pqt
  • data/processed/Qwen2.5-7B-Instruct/base/persona_compass.pqt
  • data/processed/zephyr-7b-beta/base/persona_compass.pqt
• Right-authoritarian condition files:
  • data/processed/Llama-3.1-8B-Instruct/right_authoritarian_personas/persona_compass.pqt
  • data/processed/Mistral-7B-Instruct-v0.3/right_authoritarian_personas/persona_compass.pqt
  • data/processed/Qwen2.5-7B-Instruct/right_authoritarian_personas/persona_compass.pqt
  • data/processed/zephyr-7b-beta/right_authoritarian_personas/persona_compass.pqt
• Left-libertarian condition files:
  • data/processed/Llama-3.1-8B-Instruct/left_libertarian_personas/persona_compass.pqt
  • data/processed/Mistral-7B-Instruct-v0.3/left_libertarian_personas/persona_compass.pqt
  • data/processed/Qwen2.5-7B-Instruct/left_libertarian_personas/persona_compass.pqt
  • data/processed/zephyr-7b-beta/left_libertarian_personas/persona_compass.pqt

This dataset contains synthetic programming tasks and corresponding code samples generated based on diverse, machine-created programmer personas. Unlike typical AI-generated content datasets that rely on human-written prompts or completions, this collection avoids conditioning on prior human generations, aiming to reduce bias in synthetic data creation. The methodology draws inspiration from the PersonaHub dataset. We first defined 9 key features of a programmer, including:

Primary… See the full description on the dataset page: https://huggingface.co/datasets/project-droid/DroidCollection-Personas.

TextBooksPersonaHub

  Overview

The TextBooksPersonaHub dataset is an extension of the proj-persona/PersonaHub dataset, created using the technique described in the paper Textbooks Are All You Need II. This dataset contains synthetically generated "textbook-like" passages tailored in french to specific personas, aimed at enhancing language model training with high-quality and diverse content.

  Dataset Creation







  Source Data

The original personas… See the full description on the dataset page: https://huggingface.co/datasets/drodin/TextBooksPersonaHub-FR.

Dataset Card for PersonaHub FineWeb-Edu 4 Clustering 100k

This dataset has been created with distilabel. The following figure is a map of the clusters generated from the pipeline. It's automatically generated by the TextClustering with all the information gathered. It contains 177 different clusters, which were assigned a set of 3 labels each, and the black dots correspond to those unclassified examples.

  Dataset Summary

This dataset has been… See the full description on the dataset page: https://huggingface.co/datasets/argilla/FinePersonas-v0.1-clustering-100k.

Wikipedia Personas

Wikipedia Personas is a dataset constructed from paragraphs sampled from the agentlans/wikipedia-paragraphs-complete dataset using the sample_k10000 and sample_k20000 configurations. Each paragraph is paired with a persona crafted as a plausible expert, enthusiast, or stakeholder related to the content of the corresponding Wikipedia text. Personas were initially seeded with 20 handcrafted examples following the style of proj-persona/PersonaHub and then expanded… See the full description on the dataset page: https://huggingface.co/datasets/agentlans/wikipedia-personas.

Diverse Persona 10K

This dataset is a length and diversity filtered subset (5%) of proj-persona/PersonaHub.

  Filtering Details

persona with less than 5 spaces were removed to filter out the least informative personas such as "a supporter of Die Linke". This reduced the number of personas to 196k. Then, a maxmin sampling over cosine distances of text embeddings (from voyage-3-lite) were applied to select this 10k subset. Pairwise distance statistics before maxmin filtering.… See the full description on the dataset page: https://huggingface.co/datasets/ychen/diverse-persona-10k.

Synthetic Customer Experience Persona

  Overview

The Synthetic Customer Experience Persona Dataset is a large-scale synthetic corpus of customer service personas, designed to aid in the development and evaluation of AI models for customer service applications. Inspired by Tencent AI Labs' Persona Hub, this dataset provides a diverse array of customer profiles across multiple industries.

  Dataset Statistics

Total Personas: 250,000 Industries Covered: 6 (Retail… See the full description on the dataset page: https://huggingface.co/datasets/CordwainerSmith/CustomerPersonas.

mPersonas: Multilingual Persona‑Driven Conversational Dataset

  Dataset Summary

mPersonas is a multilingual open-source dataset with high-quality persona descriptions synthetically generated by DeepSeek-V3–0324. It follows a persona-driven data synthesis methodology, similar to PersonaHub.

Instances: 510,000 Total tokens: 173M
28M in personas
145M in conversations (105M in assistant turns)

Languages: 15 License: Apache 2.0

  Methodology

This section… See the full description on the dataset page: https://huggingface.co/datasets/BSC-LT/m-personas.

How to use

from datasets import load_dataset

ds = load_dataset("jaeyong2/persona-inst", split="train") ds Dataset({ features: ['Level', 'English', 'Korean', 'Thai', 'Vietnamese', 'context'], num_rows: 3006572 })

  Development Process

Generate persona pair from proj-persona/PersonaHub We used Qwen/Qwen2-72B-Instruct model to generate Question.

  License

Qwen/Qwen2.5-72B-Instruct :… See the full description on the dataset page: https://huggingface.co/datasets/jaeyong2/persona-inst.

How to use

from datasets import load_dataset

ds = load_dataset("jaeyong2/ko-persona-cot-inst", split="train") ds Dataset({ features: ['content', 'text'], num_rows: 240000 })

  Development Process

load Question dataset from jaeyong2/persona-inst We used Qwen/Qwen2-72B-Instruct model to generate answer with COT.

  License

Qwen/Qwen2.5-72B-Instruct : https://huggingface.co/Qwen/Qwen2-72B-Instruct/blob/main/LICENSE proj-persona/PersonaHub… See the full description on the dataset page: https://huggingface.co/datasets/jaeyong2/ko-persona-cot-inst.