Search papers, labs, and topics across Lattice.
Kairos introduces a novel world model stack that transitions from passive visual generation to an operational framework for Physical AI, addressing the need for persistent state maintenance and efficient execution. It employs a Native Pre-training Paradigm with a Cross-Embodiment Data Curriculum to systematically learn from diverse data sources, while its Native Unified Architecture integrates Hybrid Linear Temporal Attention to ensure accurate state propagation over long time horizons. Experimental results demonstrate that Kairos not only achieves state-of-the-art performance on various benchmarks but also optimizes the efficiency-capability trade-off, making it a robust foundation for self-evolving physical intelligence.
Kairos achieves top-tier performance in Physical AI while ensuring low-latency execution on consumer-grade hardware, redefining the efficiency-capability landscape.
World models are transitioning from passive visual generators to foundational, operational infrastructure for Physical AI: they must natively acquire world knowledge from heterogeneous experience, maintain persistent states over long horizons, and execute efficiently within real deployment constraints. We introduce Kairos, a native world model stack designed around these requirements. (1) Kairos learns the world by pioneering a Native Pre-training Paradigm governed by a Cross-Embodiment Data Curriculum, which organizes open-world videos, human behavioral data, and robot interactions into a progressive developmental pathway. (2) Kairos maintains the world by unified world understanding, generation, and prediction within a Native Unified Architecture equipped with Hybrid Linear Temporal Attention, where sliding-window attention captures local dynamics, dilated sliding windows capture mid-range dependencies, and gated linear attention maintains persistent global memory. We establish formal theoretical bounds demonstrating that this temporal factorization strictly limits error accumulation, mathematically guaranteeing state propagation across extended horizons. (3) Kairos runs the world by incorporating a Deployment-Aware System Co-Design to support low-latency rollout generation on server and consumer-grade hardware for real-world observation-action-feedback loops. Experiments on embodied world-model, long-horizon, and action-policy benchmarks show that Kairos achieves top level performance while offering a strong efficiency-capability trade-off. Together, these results position Kairos as a cohesive operational foundation for future self-evolving physical intelligence.
