Hao Wang · page 2

27 papers in our library · 2015–2026 · page 2 of 2

Papers

The most cited first (OpenAlex and our library).

SafetyMem: Adaptive Jailbreak Defense via Dual-Component Safety Memory (2026)
LLM jailbreak defenses tend to fall into two camps. Methods that change model weights, such as fine-tuning or trained detectors, are costly and slow to adapt. Inference-time filters are flexible, but they start from scratch on every prompt. SafetyMem is a training-free framework that lets a model accumulate defense…
SOLAR-RL: Semi-Online Long-horizon Assignment Reinforcement Learning (2026)
The paper targets long-horizon navigation for GUI agents built on MLLMs. Offline RL on step-level data ignores trajectory-wide outcomes, while online RL is costly and unstable on workflows of 30+ steps. Both suffer from the credit assignment problem. SOLAR-RL takes a semi-online route: it simulates online feedback…