On the Relaxation of Conditional Independence Assumption for Image Segmentation
Spotlight (Top 3.7%)
I am a PhD student in the Department of Statistics and Data Science at The Chinese University of Hong Kong, supervised by Prof. Ben Dai. Previously, I obtained my Bachelor's degree from Beihang University.
I have broad research interests in machine learning. My earlier work centered on developing ML algorithms with statistical consistency: given a task and the metric it will be evaluated on, how a principled co-design of the loss function and the inference algorithm can optimize that metric directly rather than through a heuristic surrogate.
More recently, my interests have shifted to large language models and agentic AI, including data synthesis, benchmarking, harness engineering, and post-training.
Spotlight (Top 3.7%)
A segmentation algorithm that directly optimizes Dice and IoU with theoretical guarantees. Grounded in Bayesian decision theory, training-free, and applied at inference within three-lines code.