WARD is a guard model trained on 177K web samples and adversarially hardened via attacker-guard co-evolution to achieve high recall on prompt injections with low false positives and no added latency.
Eta: Evaluating then aligning safety of vision language models at inference time.arXiv preprint arXiv:2410.06625, 2024
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 2roles
background 1polarities
background 1representative citing papers
A neuron-targeted safety tuning method for VLLMs reduces attack success rates from ~20-30% to ~4-6% on average across ten languages while using less than 0.03% of parameters.
citing papers explorer
-
WARD: Adversarially Robust Defense of Web Agents Against Prompt Injections
WARD is a guard model trained on 177K web samples and adversarially hardened via attacker-guard co-evolution to achieve high recall on prompt injections with low false positives and no added latency.
-
Targeted Interpretable Safety Neuron Enhancement for Multilingual Vision-Language Large Models
A neuron-targeted safety tuning method for VLLMs reduces attack success rates from ~20-30% to ~4-6% on average across ten languages while using less than 0.03% of parameters.