Yi Yang

BioStance reveals that nuanced bioethical discussions on social media can be effectively modeled with high reliability, challenging the limitations of existing stance detection datasets.

SpecLoR cuts through noise and drift in text-to-video generation, achieving remarkable motion coherence with minimal computational cost.

Rejecting tokens isn't always the end—VIA-SD's slim-verifier can accurately validate many candidates, slashing inference costs and boosting efficiency.

A compact 1D token interface can dramatically enhance multimodal image fusion, achieving superior global coherence without sacrificing local detail.