JANUS: A Lightweight Framework for Jailbreaking Text-to-Image Models via Distribution Optimization

Text-to-image (T2I) models such as Stable Diffusion and DALLE remain susceptible to generating harmful or Not-Safe-For-Work (NSFW) content under jailbreak attacks despite deployed safety filters. Existing jailbreak attacks either rely on proxy-loss optimization instead of the true end-to-end objective, or depend on large-scale and costly RL-trained generators. Motivated by these limitations, we propose JANUS , a lightweight framework that formulates jailbreak as optimizing a structured prompt distribution under a black-box, end-to-end reward from the T2I system and its safety filters. JANUS replaces a high-capacity generator with a low-dimensional mixing policy over two semantically anchored prompt distributions, enabling efficient exploration while preserving the target semantics. On modern T2I models, we outperform state-of-the-art jailbreak methods, improving ASR-8 from 25.30% to 43.15% on Stable Diffusion 3.5 Large Turbo with consistently higher CLIP and NSFW scores. JANUS succeeds across both open-source and commercial models. These findings expose structural weaknesses in current T2I safety pipelines and motivate stronger, distribution-aware defenses. Warning: This paper contains model outputs that may be offensive.

Explainable Token-level Noise Filtering for LLM Fine-tuning Datasets

Large Language Models (LLMs) have seen remarkable advancements, achieving state-of-the-art results in diverse applications. Fine-tuning, an important step for adapting LLMs to specific downstream tasks, typically involves further training on corresponding datasets. However, a fundamental discrepancy exists between current fine-tuning datasets and the token-level optimization mechanism of LLMs: most datasets are designed at the sentence-level, which introduces token-level noise, causing negative influence to final performance. In this paper, we propose XTF, an explainable token-level noise filtering framework. XTF decomposes the complex and subtle contributions of token-level data to the fine-tuning process into three distinct and explicit attributes (reasoning importance, knowledge novelty, and task relevance), which can be assessed using scoring methods, and then masks the gradients of selected noisy tokens accordingly to optimize the performance of fine-tuned LLMs. We conduct extensive experiments on three representative downstream tasks (math, code and medicine) across 7 mainstream LLMs. The results demonstrate that XTF can significantly improve downstream performance by up to 13.7% compared to regular fine-tuning. Our work highlights the importance of token-level dataset optimization, and demonstrates the potential of strategies based on attribute decomposition for explaining complex training mechanisms.

Discriminative Estimation of Total Variation Distance: A Fidelity Auditor for Generative Data

With the proliferation of generative AI and the increasing volume of generative data (also called as synthetic data), assessing the fidelity of generative data has become a critical concern. In this paper, we propose a discriminative approach to estimate the total variation (TV) distance between two distributions as an effective measure of generative data fidelity. Our method quantitatively characterizes the relation between the Bayes risk in classifying two distributions and their TV distance. Therefore, the estimation of total variation distance reduces to that of the Bayes risk. In particular, this paper establishes theoretical results regarding the convergence rate of the estimation error of TV distance between two Gaussian distributions. We demonstrate that, with a specific choice of hypothesis class in classification, a fast convergence rate in estimating the TV distance can be achieved. Specifically, the estimation accuracy of the TV distance is proven to inherently depend on the separation of two Gaussian distributions: smaller estimation errors are achieved when the two Gaussian distributions are farther apart. This phenomenon is also validated empirically through extensive simulations. In the end, we apply this discriminative estimation method to rank fidelity of synthetic image data using the MNIST dataset.