4 papers
The Premise Is the Problem: Exchangeability Failure in Self-Monitored Test-Time Adaptation
Weijia Han, Lisha Qu, Zhenda Li +1
Modern forecasting models are often updated after deployment so they can respond to changing data. These updates can also make predictions worse, so practical systems need a reliab…
When the Martingale Never Stops Firing: Anytime-Valid Gating on Real Forecast Streams
Weijia Han, Lisha Qu
Machine learning systems are increasingly corrected while they run, and the decision of when to intervene is increasingly delegated to statistical monitors. Anytime-valid inference…
The Label Complexity of Useful Class-Conditional Prediction Sets under Distribution Shift
Weijia Han, Lisha Qu, Tianxin Zhou +2
Prediction sets can make deployed classifiers safer by returning several plausible labels when a single prediction is uncertain. Their value depends on classwise reliability: avera…
Decomposing Runtime, Kernel, and Quantization Speedups via a Matched FP16 Intermediate: A Hardware-Conditioned Case Study on Four NVIDIA RTX A5000 GPUs
Weijia Han, Lisha Qu
Reported serving speedups from quantized kernels typically bundle the weight format, the kernel, and the inference runtime into one number. We present an attribution study on four…