Wanyu Zhang
Titles as the sources gave them. In 1 entry, with 0 checked quotations.
In the Base
Every entry they appear in.
TERM: Target-Free Jailbreak Defense via Entropy-Guided Residual Modulation
According to the abstract, Yunkai Dang, Wanyu Zhang, Meiyi Zhu and three co-authors propose TERM, an inference-time defense against jailbreak prompts. It uses each prompt's attention entropy, calibrated against that prompt's own baseline, to adjust the model's residual updates while limiting refusals of benign requests.