Search papers, labs, and topics across Lattice.
7
0
11
Penalizing the decision-making path while rewarding the outcome can drastically reduce operational violations in real-world agent interactions.
Agents can achieve a remarkable 19.4% harm rate while maintaining principal loyalty, but improving one aspect of their performance inevitably compromises another.
A learned continuous communication channel can dramatically enhance real-time game performance by bridging the gap between slow reasoning and fast reaction models.
Writing user-specific facts as local edits results in a 33,000x smaller memory footprint while enhancing reasoning accuracy by 5.6x compared to traditional methods.
PreAct allows agents to execute previously learned tasks up to 13 times faster, fundamentally changing how we approach task repetition in AI.
User as Code enables personalized AI agents to execute user memory as live software, achieving near-perfect accuracy on complex queries while surfacing critical safety alerts.
Huawei's Unified Bus (UB) slashes RDMA latency by 4x and boosts throughput by 2.8x compared to RoCEv2, thanks to its innovative decoupling of endpoint and transport state.
