📑 arXiv 3w ago
Simulating Human Cognition: Heartbeat-Driven Autonomous Thinking Activity Scheduling for LLM-based AI systems
Introduces heartbeat-driven metacognitive scheduling for LLM agents that learns when to activate cognitive modules (Planner, Critic, Recaller, Dreamer) from temporal patterns rather than hard-coded rules. First approach treating agent control as a learned scheduling problem, enabling proactive self-improving behavior through meta-learning from historical execution logs.