Heng Fan

The paper introduces Tail-risk Intrinsic Geometric Smoothing (TIGS), a plug-and-play, inference-time defense that suppresses backdoor attacks on LLMs by structurally smoothing the attention mechanism without requiring model retraining or external data.

The paper introduces Token-Aware Gradient Optimization (TAGO), demonstrating that sparse optimization focusing only on high-gradient audio tokens is sufficient for effective jailbreaking of audio language models, making dense updates redundant.

LoRe is a training-free wrapper that dynamically budgets interaction evaluation at each step of graph solvers, significantly improving scalability and speed while maintaining solution quality.

AdaptR1 is a novel Reinforcement Learning framework that adaptively manages reasoning effort at every step of multi-hop Question Answering, significantly reducing unnecessary computational cost without sacrificing performance.