We propose SNAP, a novel low-latency Test-Time Adaptation framework that enables efficient model adaptation on edge devices by using sparse updates, significantly reducing computation while maintaining accuracy.

We introduce MGAudio, a flow-based framework for video-to-audio generation that leverages model-guided dual-role alignment to achieve state-of-the-art performance.

In the field of predictable time series forecasting task, the newly identified issue of output alignment, the metrics to evaluate it, and potential solutions.