This paper proposes Agent-Centric Interactive World Proxies, reframing world modeling from predicting physical state transitions toward supplying information that agents can directly use. Based on the provided abstract, it organizes proxies into six functional forms: dynamics, spatial, execution, memory/experience, skill, and reward/verification. It also describes three levels of agent improvement: inference-time guidance, training-time optimization, and agent-proxy co-evolution. The contribution appears primarily conceptual, offering a taxonomy and research roadmap rather than reporting a new model or verified benchmark result.
No heat snapshots are available in the last 24 hours.