Chenyang Zhao

Forget prompt engineering – LenVM lets you directly control LLM generation length with token-level precision, outperforming closed-source models on exact length matching.

Text-based speculative decoding falls flat for vision-language models, but ViSkip dynamically adapts to vision tokens for state-of-the-art acceleration.

LLMs can guide multi-objective reinforcement learning to achieve balanced trade-offs in complex systems like urban traffic control, sidestepping the need for hand-engineered reward functions.