Scholarlink

In the domain of open-ended RL, the focus is on the unsupervised discovery of skills, which is crucial for developing AGI. Mujika et al. introduce an iterative process that creates pairs of neural reward functions and policies, enabling learning diverse and complex skills in high-dimensional robotics environments without relying on feature engineering.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

From Thought to Action: Bridging Cognitive Processes and Autonomous MORL Towards Intelligent Agents in a Virtual Environment

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Shagofta Shabashkhan,
  • Xiaoyang Wang,
  • Cédric S. Mesnage

摘要

In the domain of open-ended RL, the focus is on the unsupervised discovery of skills, which is crucial for developing AGI. Mujika et al. introduce an iterative process that creates pairs of neural reward functions and policies, enabling learning diverse and complex skills in high-dimensional robotics environments without relying on feature engineering.