Hierarchical Safety Realignment: Lightweight Restoration of Safety in Pruned Large Vision-Language Models
This addresses safety concerns for deploying pruned models in resource-constrained environments, representing an incremental advance in model compression.
The paper tackles the problem of safety degradation in pruned large vision-language models by introducing Hierarchical Safety Realignment (HSR), a lightweight method that restores safety by selectively targeting critical attention heads and neurons, achieving notable improvements across various models and pruning strategies.
With the increasing size of Large Vision-Language Models (LVLMs), network pruning techniques aimed at compressing models for deployment in resource-constrained environments have garnered significant attention. However, we observe that pruning often leads to a degradation in safety performance. To address this issue, we present a novel and lightweight approach, termed Hierarchical Safety Realignment (HSR). HSR operates by first quantifying the contribution of each attention head to safety, identifying the most critical ones, and then selectively restoring neurons directly within these attention heads that play a pivotal role in maintaining safety. This process hierarchically realigns the safety of pruned LVLMs, progressing from the attention head level to the neuron level. We validate HSR across various models and pruning strategies, consistently achieving notable improvements in safety performance. To our knowledge, this is the first work explicitly focused on restoring safety in LVLMs post-pruning.