Do VLMs Share Safety Neurons Across Modalities?
A causal, neuron-level analysis of safety mechanisms in 10 VLMs, a two-stage detection pipeline with iterative ablation that accounts for self-repair, and two modality-isolated benchmarks, ViSafe-Detect and ViSafe-Eval, which decouple visual and textual safety signals are introduced.
Jia-Xuan Li, Jia-Hao Zhang, D. Vo et al.
· 0 citations