Haokai Ma

SnapGuard proposes a lightweight, multimodal method to detect prompt injection attacks in screenshot-based web agents by analyzing visual stability and contrast-polarity textual signals, achieving high accuracy with significantly reduced computational overhead.

The paper proposes TRACE, a trajectory risk-aware compression method, to effectively aggregate sparse and delayed safety evidence across long agent trajectories, achieving state-of-the-art performance on multiple safety benchmarks.

ResMerge proposes a residual-based spectral merging framework that improves the combination of multiple reinforcement learning (RL) expert models by stabilizing the aggregation process using a residual backbone.