thuml/webarena-world-model-cot

---
license: mit
tags:
- web-agent
- webarena
- world-model
---
See https://github.com/thuml/RLVR-World for examples for using this dataset.

## Citation

```
@article{wu2025rlvr,
title={RLVR-World: Training World Models with Reinforcement Learning},
author={Jialong Wu and Shaofeng Yin and Ningya Feng and Mingsheng Long},
journal={arXiv preprint arXiv:2505.13934},
year={2025},
}

RLVR-World dataset is a collection of environments or scenarios for training world models with reinforcement learning. It likely includes data for both training and testing purposes, allowing models to predict and simulate behaviors within the environment through reinforcement learning.

搜集汇总

背景与挑战

背景概述

该数据集由清华大学THUML团队发布，专注于网页代理和世界模型任务，用于训练和评估智能代理预测网页交互后的下一个状态。数据集包含约6,410个样本，涵盖网页动作序列、可访问性树状态和链式思考推理，支持强化学习研究，并基于论文《RLVR-World: Training World Models with Reinforcement Learning》。