Tian Lan

Critical immersions of even-dimensional manifolds into $\mathbb R^m$ are proven to be real-analytic, extending fundamental regularity results for geometric variational problems.

Shifting the defense from visual to audio, this method effectively thwarts identity impersonation in 3D talking face generation while preserving visual quality.

ST$^2$U reduces restricted-knowledge re-entry by up to 45% while preserving model capabilities, reshaping how we approach test-time unlearning in language models.

Achieving perfect decision-making accuracy with just three memory states reveals a groundbreaking efficiency in managing partial observability in reinforcement learning.

Persistent homology reveals a structured approach to understanding and steering LLM responses to ambiguous questions, boosting accuracy and acceptable response rates significantly.

LLMs and VLMs excel at conceptual tasks but falter in practical execution, achieving only 42.2% success in script generation for VLSI design.