Visual Token Compression Enhances Robustness of MLLMs

Researchers show for the first time that visual token pruning enhances the robustness of Multimodal Large Language Models (MLLMs), achieving an average improvement of 13.29% in defending jailbreak attacks while also reducing inference cost. The method identifies and prunes out-of-distribution visual tokens that act as vulnerabilities, evaluated across seven benchmarks including MME.

arXiv:2607.22716v1 Announce Type: new Abstract: In this paper, we show for the first time that visual token pruning enhances the robustness of Multimodal Large Language Models MLLMs , mitigating vulnerabilities such as jailbreak attacks and hallucinations. Given that vision and language modalities cannot be perfectly aligned, the misaligned visual tokens might act as out-of-distribution OOD inputs, leading to unpredictable outputs and introducing potential vulnerabilities. Building on this insight, we aim to enhance model robustness against jailbreaks and hallucinations by reducing OOD visual tokens at robust-pruning layers, while also reducing inference cost as a side benefit. Specifically, we measure the distance between each visual token and the language feature space. Then, visual tokens with large distances are identified as OOD tokens, which can be iteratively pruned. To demonstrate the effectiveness of our method, we evaluate it on seven diverse popular benchmarks. Notably, our method yields an average improvement of 13.29\% in defending jailbreak attacks, consistently achieves competitive performance in mitigating hallucinations, and maintains strong results on general datasets like MME.