Yan Xia

WING achieves state-of-the-art CT synthesis from MRI and CBCT by transforming the regression target into windowed representations, significantly improving detail and accuracy.

ResearchStudio-Reel not only automates research dissemination but does so with unprecedented quality, outperforming both traditional methods and leading LLMs in aesthetic appeal and information accuracy.

Humanoid robots can complete laboratory tasks but often fail to meet the precision required for scientific validity, exposing a critical gap in current automation efforts.

Achieving comparable performance to full-precision models, BITEMBED slashes storage costs and enhances embedding efficiency with extreme low-bit quantization.

RPCL reshapes how we approach pair confidence in MECPE, achieving up to 2.83 percentage points improvement in extraction accuracy.

Monocular human motion capture can be dramatically improved by explicitly modeling high-order temporal dynamics like velocity and acceleration, leading to more realistic and less jittery movements.

Stop wasting compute on redundant retrieval calls: DynFrame learns the optimal frame sampling density *within* each temporal window, slashing context length and boosting performance on complex video understanding tasks.

Just because an AI can identify the trigger for an emotion doesn't mean it understands *why* that emotion occurred – and current benchmarks are misleading us.

Achieve state-of-the-art 3D medical image generation by reformulating deterministic prediction as a multi-objective drifting problem, outperforming GANs, flow-matching, and SDEs in fidelity, realism, and efficiency.