Precise Shield: Explaining and Aligning VLLM Safety via Neuron-Level Guidance

TL;DR

Precise Shield introduces a neuron-level approach to enhance VLLM safety by identifying and constraining critical safety neurons, improving robustness against multimodal and multilingual attacks while maintaining generalization.

Contribution

The paper presents a novel two-stage framework that locates safety neurons and constrains updates within this subspace, enabling transfer and preservation of safety across languages and modalities.

Findings

Safety neurons are shared across languages and modalities.

Gradient masking affects fewer than 0.03% of parameters.

The method improves safety without sacrificing multilingual and multimodal performance.

Abstract

In real-world deployments, Vision-Language Large Models (VLLMs) face critical challenges from multilingual and multimodal composite attacks: harmful images paired with low-resource language texts can easily bypass defenses designed for high-resource language scenarios, exposing structural blind spots in current cross-lingual and cross-modal safety methods. This raises a mechanistic question: where is safety capability instantiated within the model, and how is it distributed across languages and modalities? Prior studies on pure-text LLMs have identified cross-lingual shared safety neurons, suggesting that safety may be governed by a small subset of critical neurons. Leveraging this insight, we propose Precise Shield, a two-stage framework that first identifies safety neurons by contrasting activation patterns between harmful and benign inputs, and then constrains parameter updates strictly within this subspace via gradient masking with affecting fewer than 0.03% of parameters. This strategy substantially improves safety while preserving multilingual and multimodal generalization. Further analysis reveals a moderate overlap of safety neurons across languages and modalities, enabling zero-shot cross-lingual and cross-modal transfer of safety capabilities, and offering a new direction for neuron-level, transfer-based safety enhancement.