State-Text Pairs
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/AnonymousUser360/CLSP
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了大约55万个状态-文本对，这些对是从一个多人第一人称射击游戏环境中生成的，用于训练对比学习模型。其中，5%的状态-文本对被随机抽取作为测试数据集，而剩余的95%则用于训练。数据规模为55万个状态-文本对（从18分钟的数据中抽取）。所涉及的任务包括状态-文本对齐和强化学习导航。
This dataset contains approximately 550,000 state-text pairs generated from a multiplayer first-person shooter (FPS) game environment, intended for training contrastive learning models. Five percent of these state-text pairs are randomly sampled as the test dataset, while the remaining 95% are used for model training. The dataset totals 550,000 state-text pairs extracted from 18 minutes of gameplay data. The relevant tasks include state-text alignment and reinforcement learning navigation.
提供机构：
Anonymous User
