Large Language Models (LLMs) are vulnerable to attacks like prompt injection, backdoor attacks, and adversarial attacks, which manipulate prompts or models to generate harmful outputs. In this paper, departing from traditional deep learning attack paradigms, we explore their intrinsic relationship and collectively term them Prompt Trigger Attacks (PTA). This raises a key question: Given a prompt, can we tell whether a hidden trigger is steering the model’s behavior? To address this, we propose UniGuardian, the first unified, training-free defense that detects trigger-activated prompt injection, backdoor, and adversarial attacks by measuring how structured perturbations of the prompt shift the model’s output distribution. Additionally, we introduce a single-forward strategy to optimize the detection pipeline, enabling simultaneous attack detection and text generation within a single forward pass. Our experiments confirm that UniGuardian accurately and efficiently identifies trigger-activated prompts in LLMs.

Autoregressive (AR) models have recently shown strong performance in image generation, where a key component is the visual tokenizer (VT) that maps continuous pixels into discrete tokens. The quality of the VT largely determines the upper bound of AR model performance. However, existing discrete VTs significantly lag behind continuous variational autoencoders (VAEs), leading to degraded reconstructions and poor preservation of fine details and text. Current benchmarks primarily evaluate end-to-end generation quality, making it difficult to isolate VT performance. We introduce VTBench, a comprehensive benchmark for systematic evaluation of visual tokenizers across three core tasks: image reconstruction, detail preservation, and text preservation. Evaluating state-of-the-art VTs with multiple metrics, we find that continuous VAEs consistently outperform discrete VTs, especially in retaining spatial structure and semantic details. Discrete VTs often produce distorted reconstructions and fail to preserve fine textures and text. We release VTBench publicly to facilitate future research on robust, general-purpose visual tokenizers.

Considering deep neural networks as manifold mappers, the pretrain-then-fine-tune paradigm can be interpreted as a two-stage process: pretrain establishes a broad knowledge base, and fine-tune adjusts the model parameters to activate specific neural pathways to align with the target manifold. Although prior fine-tuning approaches demonstrate success, their rigid parameter space limits their ability to dynamically activate appropriate neural pathways, rendering them ill-equipped to adapt flexibly to the diverse and evolving data distributions. In light of this view, we propose a novel approach, Mixture of Expert Prompt Tuning (MEPT), as an effective and efficient manifold-mapping framework. MEPT leverages the Mixture of Experts architecture by integrating multiple prompt experts to adaptively learn diverse and non-stationary data distributions. Empirical evaluations demonstrate that MEPT outperforms several state-of-the-art parameter efficient baselines on SuperGLUE, achieving notable improvements in mean accuracy (e.g., 1.94%) while significantly reducing activated prompts by 79.25%. The effectiveness of MEPT is further supported by theoretical insights from manifold learning and validated through neural activation pathway visualization results. Our code is avaliable at https://runjia.tech/emnlp_mept/.