Mingxuan Yuan

Memory retention strategies that account for long-term consequences can dramatically enhance the performance of language agents under tight constraints.

YouZhi-LLM achieves unprecedented concurrency and accuracy in financial LLMs by dramatically reducing KV-cache overhead, setting a new standard for deployment efficiency.

LLMs can generate Verilog that's actually useful for hardware design when guided by a closed-loop system that incorporates feedback from simulation, synthesis, and timing analysis.