Search papers, labs, and topics across Lattice.
City University of Hong Kong
3
0
6
Visual verification can dramatically enhance the performance of GUI agents, revealing that traditional text-based methods fall short in long-horizon tasks.
RAG systems are surprisingly vulnerable: a tiny amount of carefully crafted, almost undetectable poisoned data can reliably hijack retrieval and manipulate generated answers.
Forensic vision-language models can be fooled into not only misclassifying tampered images, but also providing plausible, yet incorrect, explanations, highlighting a critical consistency failure.
