Despite the integration of safety alignment and external filters, text-to-image (T2I) generative systems are still susceptible to producing harmful content, such as sexual or violent imagery. This raises serious concerns about unintended exposure and potential misuse. Red teaming, which aims to proactively identify diverse prompts that can elicit unsafe outputs from the T2I system, is increasingly recognized as an essential method for assessing and improving safety before real-world deployment. However, existing automated red teaming approaches often treat prompt discovery as an isolated, prompt-level optimization task, which limits their scalability, diversity, and overall effectiveness. To bridge this gap, in this paper, we propose DREAM, a scalable red teaming framework to automatically uncover diverse problematic prompts from a given T2I system. Unlike prior work that optimizes prompts individually, DREAM directly models the probabilistic distribution of the target system's problematic prompts, which enables explicit optimization over both effectiveness and diversity, and allows efficient large-scale sampling after training. To achieve this without direct access to representative training samples, we draw inspiration from energy-based models and reformulate the objective into a simple and tractable form. We further introduce GC-SPSA, an efficient optimization algorithm that provides stable gradient estimates through the long and potentially non-differentiable T2I pipeline. During inference, we also propose a diversity-aware sampling strategy to enhance prompt variety. The effectiveness of DREAM is validated through extensive experiments, demonstrating state-of-the-art performance across a wide range of T2I models and safety filters in terms of both prompt success rate and diversity. Our code is available at https://github.com/AntigoneRandy/DREAM

Text-to-image (T2I) diffusion models have achieved impressive image generation quality and are increasingly fine-tuned for personalized applications. However, these models often inherit unsafe behaviors from toxic pretraining data, raising growing safety concerns. While recent safety-driven unlearning methods have made promising progress in suppressing model toxicity, they are found to be fragile to downstream fine-tuning, as we reveal that state-of-the-art methods largely fail to retain their effectiveness even when fine-tuned on entirely benign datasets. To mitigate this problem, in this paper, we propose ResAlign, a safety-driven unlearning framework with enhanced resilience against downstream fine-tuning. By modeling downstream fine-tuning as an implicit optimization problem with a Moreau envelope-based reformulation, ResAlign enables efficient gradient estimation to minimize the recovery of harmful behaviors. Additionally, a meta-learning strategy is proposed to simulate a diverse distribution of fine-tuning scenarios to improve generalization. Extensive experiments across a wide range of datasets, fine-tuning methods, and configurations demonstrate that ResAlign consistently outperforms prior unlearning approaches in retaining safety, while effectively preserving benign generation capability. Our code and pretrained models are publicly available at https://github.com/AntigoneRandy/ResAlign.

The Endless Tuning is a design method for a reliable deployment of artificial intelligence based on a double mirroring process, which pursues both the goals of avoiding human replacement and filling the so-called responsibility gap (Matthias 2004). Originally depicted in (Fabris et al. 2024) and ensuing the relational approach urged therein, it was then actualized in a protocol, implemented in three prototypical applications regarding decision-making processes (respectively: loan granting, pneumonia diagnosis, and art style recognition) and tested with such as many domain experts. Step by step illustrating the protocol, giving insights concretely showing a different voice (Gilligan 1993) in the ethics of artificial intelligence, a philosophical account of technical choices (e.g., a reversed and hermeneutic deployment of XAI algorithms) will be provided in the present study together with the results of the experiments, focusing on user experience rather than statistical accuracy. Even thoroughly employing deep learning models, full control was perceived by the interviewees in the decision-making setting, while it appeared that a bridge can be built between accountability and liability in case of damage.

Transformer models rely heavily on the scaled dot-product attention (SDPA) operation, typically implemented as FlashAttention. Characterized by its frequent interleaving of matrix multiplications and softmax operations, FlashAttention fails to fully utilize the compute resources of modern systolic-array-based accelerators designed for consecutive and large matrix multiplications. To fully unleash the performance potential of systolic arrays for FlashAttention, we propose FSA, an enhanced systolic array architecture that runs the entire FlashAttention on the array without external vector units. Combined with SystolicAttention, an optimized kernel for FSA that achieves fine-grained and element-wise overlapping of FlashAttention operations, FSA maximizes array utilization while preserving the original floating-point operation order of FlashAttention. We implement FSA in synthesizable RTL and evaluate its performance against state-of-the-art systolic-array-based accelerators. Our results show that FSA achieves 1.77x and 4.83x higher attention FLOPs/s utilization compared to AWS Neuron-v2 and Google TPUv5e, respectively. We synthesize FSA in a 16 nm technology at 1.5 GHz, and results indicate only a 12% area overhead compared to a standard weight-stationary systolic array.

Artificial intelligence (AI)-driven models have the potential to revolutionize weather forecasting, but still rely on initial conditions generated by costly Numerical Weather Prediction (NWP) systems. Although recent end-to-end forecasting models attempt to bypass NWP systems, these methods lack scalable assimilation of new types of observational data. Here, we introduce XiChen, an observation-scalable fully AI-driven global weather forecasting system, wherein the entire pipeline, from Data Assimilation (DA) to medium-range forecasting, can be accomplished within only 15 seconds. XiChen is built upon a foundation model that is pre-trained for weather forecasting and subsequently fine-tuned to serve as both observation operators and DA models, thereby enabling the scalable assimilation of conventional and raw satellite observations. Furthermore, the integration of Four-Dimensional Variational (4DVar) knowledge ensures XiChen to achieve DA and medium-range forecasting accuracy comparable to operational NWP systems, with skillful forecasting lead time beyond 8.75 days. A key feature of XiChen is its ability to maintain physical balance constraints during DA, enabling observed variables to correct unobserved ones effectively. In single-point perturbation DA experiments, XiChen exhibits flow-dependent characteristics similar to those of traditional 4DVar systems. These results demonstrate that XiChen holds strong potential for fully AI-driven weather forecasting independent of NWP systems.

We introduce a generalized \textit{Probabilistic Approximate Optimization Algorithm (PAOA)}, a classical variational Monte Carlo framework that extends and formalizes prior work by Weitz \textit{et al.}~\cite{Combes_2023}, enabling parameterized and fast sampling on present-day Ising machines and probabilistic computers. PAOA operates by iteratively modifying the couplings of a network of binary stochastic units, guided by cost evaluations from independent samples. We establish a direct correspondence between derivative-free updates and the gradient of the full Markov flow over the exponentially large state space, showing that PAOA admits a principled variational formulation. Simulated annealing emerges as a limiting case under constrained parameterizations, and we implement this regime on an FPGA-based probabilistic computer with on-chip annealing to solve large 3D spin-glass problems. Benchmarking PAOA against QAOA on the canonical 26-spin Sherrington-Kirkpatrick model with matched parameters reveals superior performance for PAOA. We show that PAOA naturally extends simulated annealing by optimizing multiple temperature profiles, leading to improved performance over SA on heavy-tailed problems such as SK-Lévy.

Watermarking is a promising defense against the misuse of large language models (LLMs), yet it remains vulnerable to scrubbing and spoofing attacks. This vulnerability stems from an inherent trade-off governed by watermark window size: smaller windows resist scrubbing better but are easier to reverse-engineer, enabling low-cost statistics-based spoofing attacks. This work breaks this trade-off by introducing a novel mechanism, equivalent texture keys, where multiple tokens within a watermark window can independently support the detection. Based on the redundancy, we propose a novel watermark scheme with Sub-vocabulary decomposed Equivalent tExture Key (SEEK). It achieves a Pareto improvement, increasing the resilience against scrubbing attacks without compromising robustness to spoofing. Experiments demonstrate SEEK's superiority over prior method, yielding spoofing robustness gains of +88.2%/+92.3%/+82.0% and scrubbing robustness gains of +10.2%/+6.4%/+24.6% across diverse dataset settings.

Building on our previous work introducing Fredholm Neural Networks (Fredholm NNs/ FNNs) for solving integral equations, we extend the framework to tackle forward and inverse problems for linear and semi-linear elliptic partial differential equations. The proposed scheme consists of a deep neural network (DNN) which is designed to represent the iterative process of fixed-point iterations for the solution of elliptic PDEs using the boundary integral method within the framework of potential theory. The number of layers, weights, biases and hyperparameters are computed in an explainable manner based on the iterative scheme, and we therefore refer to this as the Potential Fredholm Neural Network (PFNN). We show that this approach ensures both accuracy and explainability, achieving small errors in the interior of the domain, and near machine-precision on the boundary. We provide a constructive proof for the consistency of the scheme and provide explicit error bounds for both the interior and boundary of the domain, reflected in the layers of the PFNN. These error bounds depend on the approximation of the boundary function and the integral discretization scheme, both of which directly correspond to components of the Fredholm NN architecture. In this way, we provide an explainable scheme that explicitly respects the boundary conditions. We assess the performance of the proposed scheme for the solution of both the forward and inverse problem for linear and semi-linear elliptic PDEs in two dimensions.

LLM-integrated applications and agents are vulnerable to prompt injection attacks, where adversaries embed malicious instructions within seemingly benign input data to manipulate the LLM's intended behavior. Recent defenses based on known-answer detection (KAD) scheme have reported near-perfect performance by observing an LLM's output to classify input data as clean or contaminated. KAD attempts to repurpose the very susceptibility to prompt injection as a defensive mechanism. We formally characterize the KAD scheme and uncover a structural vulnerability that invalidates its core security premise. To exploit this fundamental vulnerability, we methodically design an adaptive attack, DataFlip. It consistently evades KAD defenses, achieving detection rates as low as $0\%$ while reliably inducing malicious behavior with a success rate of $91\%$, all without requiring white-box access to the LLM or any optimization procedures.

-- Current medical language models, adapted from large language models (LLMs), typically predict ICD code - based diagnosis from electronic health records (EHRs) because these labels are readily available. However, ICD codes do not capture the nuanced, context - rich reasoning clinicians use for diagnosis. Clinicians synthesize diverse patient data and reference clinical practice guidelines (CPGs) to make evidence - based decisions. This misalignment limits the clinical utility of existing models. We introduce GARMLE - G, a Generation - Augmented Retrieval framework that grounds medical language model outp uts in authoritative CPGs. Unlike conventional Retrieval - Augmented Generation based approaches, GARMLE - G enables hallucination - free outputs by directly retrieving authoritative guideline content without relying on model - generated text. It (1) integrates LLM predictions with EHR data to create semantically rich queries, (2) retrieves relevant CPG knowledge snippets via embedding similarity, and (3) fuses guideline content with model output to generate clinically aligned recommendations. A prototype system for hypertension diagnosis was developed and evaluated on multiple metrics, demonstrating superior retrieval precision, semantic relevance, and clinical guideline adherence compared to RAG - based baselines, while maintaining a lightweight architecture suitable for localized healthcare deployment. This work provides a scalable, low - cost, and hallucination - free method for grounding medical language models in evidence - based clinical practice, with strong potential for broader clinical deployment. The research reported in this paper is financially supported by the National Natural Science Foundation of China (62276156), the project of Shandong Provincial Natural Science Foundation (ZR2024LZH005), the Taishan Scholar Program of Shandong Province of China (No.tsq nz20240809), and the Excellent Youth Foundation of Shandong Natural Science Foundation (2024HWYQ - 055). Wenhao Li is with Shandong Normal University, Jinan, China, 250358 (email: lwh@sdnu.edu.cn) Hongkuan Zhang is with Shandong Normal University, Jinan, China, 250358 (email: 2024217028@stu.sdnu.edu.cn) In the healthcare sector, language models and related tools, such as ChatGPT and ClinicalBERT, have been increasingly applied across multiple scenarios, including disease prediction, clinical decision support, patient interaction, drug discovery, and personalized medicine, significantly driving innovation and transformation in medical technology [1, 2] . As a fundamental task in healthcare, disease diagnosis refers to the process by which health professionals identify the most likely disease or disorder causing a patient's symptoms [3] .