Search papers, labs, and topics across Lattice.
5
0
9
2
"Late-blooming" tokens can be crucial for deep-layer reasoning, and our method ensures they aren't discarded prematurely, achieving over 77.8% token reduction without sacrificing performance.
SAGE transforms failure recovery in autonomous research by generating multiple grounded explanations, leading to a dramatic increase in reliable scientific outputs.
Achieving state-of-the-art performance in long-horizon video world models, CaR flexibly retrieves memory while minimizing computational costs.
EvolveNav achieves a 10.1% boost in success rates for zero-shot navigation by enabling agents to learn and adapt in real-time from their experiences.
Generate cinematic videos with precise narrative control by conditioning on multiple keyframes, unlocking richer storytelling than text- or single-frame-guided approaches.
