arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2310.06648cs.LGcs.AIcs.NE
收藏

来自人类反馈的多样性

Diversity from Human Feedback

National Key Laboratory for Novel Software Technology, Nanjing University(新型软件技术国家实验室);School of Artificial Intelligence, Nanjing University(人工智能学院);Department of Statistics and Data Science, Southern University of Science and Technology(统计与数据科学系);Tencent AI Lab(腾讯人工智能实验室)

Ren-Jian Wang, Ke Xue, Yutong Wang, Peng Yang, Haobo Fu, Qiang Fu, Chao Qian

AI总结:

针对多样性度量中行为空间依赖专家设计的难题,本文提出从人类反馈学习行为空间的 DivHF,并结合 MAP-Elites 在 QDax 上验证其更符合人类偏好且提升解的多样性。

摘要:

多样性在集成学习、强化学习和组合优化等诸多问题中发挥着重要作用。如何定义多样性度量是一个长期存在的问题。许多方法依赖专家经验来定义合适的行为空间,进而获得多样性度量,但这在许多场景中颇具挑战。本文提出了从人类反馈中学习行为空间的问题,并给出了一种名为 Diversity from Human Feedback(DivHF,来自人类反馈的多样性)的通用方法来解决该问题。DivHF 通过查询人类反馈,学习与人类偏好一致的行为描述符。所学行为描述符可与任意距离度量结合,以定义多样性度量。我们将 DivHF 与 Quality-Diversity(质量—多样性)优化算法 MAP-Elites 集成,并在 QDax 套件上开展实验,验证了其有效性。结果表明,与没有人类反馈的直接数据驱动方法所学行为相比,DivHF 所学行为与人类需求一致得多,并使最终解在人类偏好下更加多样。我们的贡献包括形式化该问题、提出 DivHF 方法,以及通过实验证明其有效性。

英文摘要:

Diversity plays a significant role in many problems, such as ensemble learning, reinforcement learning, and combinatorial optimization. How to define the diversity measure is a longstanding problem. Many methods rely on expert experience to define a proper behavior space and then obtain the diversity measure, which is, however, challenging in many scenarios. In this paper, we propose the problem of learning a behavior space from human feedback and present a general method called Diversity from Human Feedback (DivHF) to solve it. DivHF learns a behavior descriptor consistent with human preference by querying human feedback. The learned behavior descriptor can be combined with any distance measure to define a diversity measure. We demonstrate the effectiveness of DivHF by integrating it with the Quality-Diversity optimization algorithm MAP-Elites and conducting experiments on the QDax suite. The results show that the behavior learned by DivHF is much more consistent with human requirements than the one learned by direct data-driven approaches without human feedback, and makes the final solutions more diverse under human preference. Our contributions include formulating the problem, proposing the DivHF method, and demonstrating its effectiveness through experiments.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑