Rich Observation Combination Lock (comblock)

该数据集是一个基准环境，它要求从丰富的观察中恢复正确的特征，并进行策略性探索，以此来测试强化学习中表征转移的有效性。数据集涉及策略性探索和潜在状态的发现，重点关注源任务中生成性访问的影响。其规模包括多个源任务及一个目标任务，所涉及的任务领域为强化学习。

This dataset is a benchmark environment that requires recovering correct features from rich observations and conducting strategic exploration to test the effectiveness of representation transfer in reinforcement learning. It involves strategic exploration and the discovery of latent states, with a focus on the impact of generative access in source tasks. It encompasses multiple source tasks and one target task, and the involved task domain is reinforcement learning.