Jiqiang Liu

Erasing and Refining Discriminative Features for CNV Subtype Classification in OCT Images

Choroidal neovascularization (CNV) subtype classification from optical coherence tomography (OCT) images is clinically important because treatment response and disease prognosis vary across subtypes. However, automated classification remains challenging owing to subtle inter-class differences and considerable imaging n...

Class transfer estimation in neural networks for backdoor target identification and mitigation

MileGPO: Milestone Inference with Local Evidence for Graph-Based Policy Optimization of Long-Horizon LLM Agents

MileGPO (Milestone Inference with Local Evidence for Graph-Based Policy Optimization) is proposed, which derives process-level credit from grouped on-policy rollouts through three designs and indicates that reliability weighting, local progress, and same-state branch evidence complement milestone discovery and resolve...

Sem-DRL: Network-Aware Continuous Reinforcement Learning for Malicious Semantic Routing in 6G Networks

The evolution of 6G networks from bit-oriented transmission to semantic communication renders communication systems vulnerable to emerging security threats targeting information interpretation, specifically semantic distortion. Consequently, identifying critical attack paths is essential for understanding and mitigatin...

Embodied Agentic Intelligence for LAWNs: Joint Collision Avoidance Path Planning and Predictive Beamforming Design

Low-altitude wireless networks (LAWNs) are emerging as a foundational infrastructure for the low-altitude economy, necessitating unmanned aerial vehicles (UAVs) to operate as embodied wireless agents that can navigate complex air-ground environments. This paper investigates the critical challenge of maintaining reliabl...

Mask2Keep: Mask-guided information transfer for backdoors resilient to compression-oriented pruning.

Mask2Keep (M2K), an attack-enhancement training framework that improves backdoor robustness under compression-oriented pruning by encouraging the model to encode both task- and backdoor-related information into parameters that are less likely to be removed, is proposed.