MLLMs are increasingly deployed in user-facing applications, yet they inherit backdoor risks from the pipelines used to construct them: triggers may reside in images, texts, or both. Existing model-level backdoor removal methods, largely designed for conventional classifiers, show limited effectiveness on MLLMs, while...
Jia-Li Wei, Ming Fan, Ming-Kun Zhang et al.· 0 citations
This work formalizes influence as a counterfactual estimand, distinguishes specification mismatch across estimands from approximation error in estimating a fixed estimand, and organizes representative estimators by their implied specifications.
Zhe Li, Wei Zhao, Pei-Xin Zhang et al.· 0 citations
This work constructs a cross-effect matrix over a taxonomy of 67 behaviors across three open-weight language models and shows that side effects are largely predictable before steering is performed, enabling proactive safety auditing and more informed deployment of steering interventions.
Yong-Ong Chong, Alson Wei Jie Sim, Pei-Xin Zhang et al.· 1 citation
A training-free defense that first identifies safety-specific neurons through per-neuron hypothesis tests under false-discovery-rate control together with a utility-specificity filter, and reduces the average attack success rate to at most 2.0% while incurring a utility drop of only 0.3% on MT-Bench, the smallest among...
Wei Zhao, Zhe Li, Pei-Xin Zhang et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
