collaborators

6 papers

cs.CV2026

A Controlled Audit of Pretraining Contamination in Public Medical Vision-Language Benchmarks

Bruce Changlong Xu, Lan Wu, Alexander Ryu

Medical vision-language models (VLMs) are evaluated on public benchmarks whose images and question-answer pairs have been freely downloadable for years, yet reported accuracy assum…

cs.LG2026

Alignment Collapse Under KV Cache Quantization: Diagnosis and Mitigation

Bruce Changlong Xu, Adarsh Kumarappan, Mu Zhou

Key-value (KV) cache quantization is widely used to reduce Large Language Model (LLM) inference memory, yet existing evaluations solely focus on measuring perplexity and accuracy w…

cs.CL2026

Typhoon: Towards an Effective Task-Specific Masking Strategy for Pre-trained Language Models

Muhammed Shahir Abdurrahman, Hashem Elezabi, Bruce Changlong Xu

The choice of \emph{which} tokens to mask is a central, under-examined design decision in masked language modeling (MLM). Standard pretraining masks tokens uniformly at random, but…

cs.LG2026

Training-Inference Kernel Contracts: Bounding Divergence in Post-Training and Deployment

Bruce Changlong Xu, Lan Wu

A modern post-training pipeline often writes one symbol for its policy, pi_theta, while evaluating it through two different programs: a training kernel optimized for autograd and a…

cs.CV2026

From Theory to Decision Rule: Calibrating the Noisy-Label Crossover for Vision-Language Model Weak Supervision Across Three Medical-Imaging Benchmarks

Bruce Changlong Xu, Jose James, Alexander Ryu

Classical noisy-label theory predicts that downstream performance under weak supervision is bounded above by the labeler's accuracy, implying a sharp crossover: once a gold-trained…

cs.LG2026

Activation Sensitivity as a Unifying Principle for Post-Training Quantization

Bruce Changlong Xu

Post-training quantization (PTQ) methods for large language models rely on heuristics that implicitly estimate which weight channels most strongly influence model behavior. Two dom…