Xiaochun Yang

For the first time, pointwise stability is achieved for 3D compressible adiabatic flows through porous media, revealing a nonlinear comparison principle that links solution differences to equilibrium state and initial perturbation distances.

Simple adaptive negative sampling can significantly boost the performance of knowledge graph foundation models on zero-shot completion tasks.

Injecting ontological knowledge into temporal knowledge graph extrapolation models can substantially boost performance, especially for sparsely connected entities.

LLMs can be made more effective at KBQA by strategically incorporating SLM-based KBC models into their reasoning process, and vice versa.