Xiaoyu Li

Conflicting findings on model representation convergence are unified by a lattice structure, revealing that "Platonic" global agreement is only one facet of a broader representational landscape.

Few labels per question are sufficient for near-optimal recalibration when information is pooled across related tasks, with identification regret decaying exponentially with label count.

Hyper-connections create a fading memory in Transformers, with the effective context horizon directly tied to the second singular value of the stream-mixing matrices.

The results establish explicit-implicit matching as an effective and efficient mechanism for spatio-temporal instance interaction and Integrating temporal MatchFusion into SparseDrive further improves perception within an E2E framework without additional supervision.