Mang Ye

Current chemistry benchmarks conflate memorization with derivation and treat visual/textual inputs interchangeably; ChemArgus finally provides fine-grained diagnostic scores that pinpoint *where* a model's reasoning chain breaks.

Animal visual re-identification no longer requires shared anatomical priors: decoupling environmental style from dynamic cross-species relational neighborhoods enables unified representation learning across radically different morphologies.

DriveCache boosts driving video generation efficiency by intelligently caching based on planned motion, leading to significant improvements in fidelity and throughput.

Monocular 3D priors can enhance person re-identification without succumbing to noise, achieving better results in challenging visual conditions.