LingoLoop Attack: Trapping MLLMs via Linguistic Context and State Entrapment into Endless Loops

2025-06-17 · Jiyuan Fu, Kaixun Jiang, Lingyi Hong, Jinglun Li, Haijing Guo, Dingkang Yang, Zhaoyu Chen, Wenqiang Zhang

Multimodal Large Language Models (MLLMs) have shown great promise but require substantial computational resources during inference. Attackers can exploit this by inducing excessive output, leading to resource exhaustion and service degradation. Prior energy-latency attacks aim to increase generation time by broadly shifting the output token distribution away from the EOS token, but they neglect the influence of token-level Part-of-Speech (POS) characteristics on EOS and sentence-level structural patterns on output counts, limiting their efficacy. To address this, we propose LingoLoop, an attack designed to induce MLLMs to generate excessively verbose and repetitive sequences. First, we find that the POS tag of a token strongly affects the likelihood of generating an EOS token. Based on this insight, we propose a POS-Aware Delay Mechanism to postpone EOS token generation by adjusting attention weights guided by POS information. Second, we identify that constraining output diversity to induce repetitive loops is effective for sustained generation. We introduce a Generative Path Pruning Mechanism that limits the magnitude of hidden states, encouraging the model to produce persistent loops. Extensive experiments demonstrate LingoLoop can increase generated tokens by up to 30 times and energy consumption by a comparable factor on models like Qwen2.5-VL-3B, consistently driving MLLMs towards their maximum generation limits. These findings expose significant MLLMs' vulnerabilities, posing challenges for their reliable deployment. The code will be released publicly following the paper's acceptance.

📄 PDF Abstract BibTeX arXiv:2506.14493

Code (0)

등록된 구현이 없습니다.

Tasks

POSTAG

Methods 이 논문이 사용한 방법론

Pruning 설명 없음

Efficient Temporal Extrapolation of Multimodal Large Language Models with Temporal Grounding Bridge

2024-02-25 · Yuxuan Wang, Yueqian Wang, Pengfei Wu, Jianxin Liang 외

Despite progress in multimodal large language models (MLLMs), the challenge of interpreting long-form videos in response to linguistic queries persists, largely due to the inefficiency in temporal grounding and limited p…

Computational Efficiency Language Modelling Optical Flow Estimation Question Answering+1

SaFeR-Steer: Evolving Multi-Turn MLLMs via Synthetic Bootstrapping and Feedback Dynamics

2026-03-18 · Haolong Hu, Hanyu Li, Tiancheng He, Huahui Yi 외 arxiv

MLLMs are increasingly deployed in multi-turn settings, where attackers can escalate unsafe intent through the evolving visual-text history and exploit long-context safety decay. Yet safety alignment is still dominated b…

Mean BERTs make erratic language teachers: the effectiveness of latent bootstrapping in low-resource settings

2023-10-30 · David Samuel

This paper explores the use of latent bootstrapping, an alternative self-supervision technique, for pretraining language models. Unlike the typical practice of using self-supervision on discrete subwords, latent bootstra…

Dynamic Adversarial Reinforcement Learning for Robust Multimodal Large Language Models

2026-01-24 · Yicheng Bao, Xuhong Wang, Qiaosheng Zhang, Chaochao Lu 외 arxiv

Despite their impressive capabilities, Multimodal Large Language Models (MLLMs) exhibit perceptual fragility when confronted with visually complex scenes. This weakness stems from a reliance on finite training datasets, …

Reinforcement Learning

Evaluating Linguistic Capabilities of Multimodal LLMs in the Lens of Few-Shot Learning

2024-07-17 · Mustafa Dogan, Ilker Kesen, Iacer Calixto, Aykut Erdem 외

The linguistic capabilities of Multimodal Large Language Models (MLLMs) are critical for their effective application across diverse tasks. This study aims to evaluate the performance of MLLMs on the VALSE benchmark, focu…

Few-Shot Learning In-Context Learning