Zhiling Wang

Retrieval-based visual reasoning can match state-of-the-art performance with half the parameters, revolutionizing how we think about multimodal learning.

History-guided control in DFP allows for adaptable motion planning that avoids the pitfalls of static pattern copying, leading to safer and more stable driving trajectories.

LLMs can be efficiently post-trained by only updating half the parameters, slashing memory costs without sacrificing performance.

Stop rewarding reasoning that just looks good – reward reasoning that actually *helps* the downstream model solve the task.