Search papers, labs, and topics across Lattice.
UIUC UIUC UIUC, Bytedance University of Edinburgh Bytedance
3
0
7
Lodestar slashes time-to-first-token by up to 4.42x in heterogeneous GPU clusters through intelligent, adaptive request routing.
Learning-based congestion controllers are surprisingly more robust to adversarial attacks than traditional algorithms, and these attacks can be used to train even better controllers.
LLMs might sound good at designing networked systems, but they're surprisingly bad at avoiding configurations that violate basic constraints, highlighting the need for structured reasoning frameworks like Kepler.
