Last week, shortly before the start of the Chinese New Year, when much of China shuts down for seven days, the state media saluted DeepSeek, a tech startup whose release of a new low-cost, high-performance artificial-intelligence model, known as R1, prompted a big sell-off in tech stocks on Wall Street. China Central Television showed footage of DeepSeek's bespectacled founder, Liang Wenfeng, meeting with Premier Li Qiang, the second-highest-ranking official in the Chinese government. A few days earlier, China Daily, an English-language news site run by the Chinese Communist Party, had hailed DeepSeek's success, which defied U.S. restrictions on the export of high-performance semiconductor chips used to train A.I. models, as "not an isolated phenomenon, but rather a reflection of the broader vibrancy of China's AI ecosystem." As if to reinforce the point, on Wednesday, the first day of the Year of the Snake, Alibaba, the Chinese tech giant, released its own new A.I. model, which the company claimed "outperforms" competing products from U.S. companies like OpenAI and Meta "almost across the board." Alibaba's claims haven't been independently verified yet, but the DeepSeek-inspired stock sell-off provoked a great deal of commentary about how the company achieved its breakthrough, the durability of U.S. leadership in A.I., and the wisdom of trying to slow down China's tech industry by restricting high-tech exports--a policy that both the first Trump Administration and the Biden Administration followed.

In Distributed Manipulator Systems (DMS), decentralization is a highly desirable property as it promotes robustness and facilitates scalability by distributing computational burden and eliminating singular points of failure. However, current DMS typically utilize a centralized approach to sensing, such as single-camera computer vision systems. This centralization poses a risk to system reliability and offers a significant limiting factor to system size. In this work, we introduce a decentralized approach for sensing and in a Distributed Manipulator Systems using Neural Cellular Automata (NCA). Demonstrating a decentralized sensing in a hardware implementation, we present a novel inductive sensor board designed for distributed sensing and evaluate its ability to estimate global object properties, such as the geometric center, through local interactions and computations. Experiments demonstrate that NCA-based sensing networks accurately estimate object position at 0.24 times the inter sensor distance. They maintain resilience under sensor faults and noise, and scale seamlessly across varying network sizes. These findings underscore the potential of local, decentralized computations to enable scalable, fault-tolerant, and noise-resilient object property estimation in DMS

As large language models (LLMs) are becoming more capable and widespread, the study of their failure cases is becoming increasingly important. Recent advances in standardizing, measuring, and scaling test-time compute suggest new methodologies for optimizing models to achieve high performance on hard tasks. In this paper, we apply these advances to the task of model jailbreaking: eliciting harmful responses from aligned LLMs. We develop an adversarial reasoning approach to automatic jailbreaking via test-time computation that achieves SOTA attack success rates (ASR) against many aligned LLMs, even the ones that aim to trade inference-time compute for adversarial robustness. Our approach introduces a new paradigm in understanding LLM vulnerabilities, laying the foundation for the development of more robust and trustworthy AI systems.

Navigating densely vegetated environments poses significant challenges for autonomous ground vehicles. Learning-based systems typically use prior and in-situ data to predict terrain traversability but often degrade in performance when encountering out-of-distribution elements caused by rapid environmental changes or novel conditions. This paper presents a novel, lidar-only, online adaptive traversability estimation (TE) method that trains a model directly on the robot using self-supervised data collected through robot-environment interaction. The proposed approach utilises a probabilistic 3D voxel representation to integrate lidar measurements and robot experience, creating a salient environmental model. To ensure computational efficiency, a sparse graph-based representation is employed to update temporarily evolving voxel distributions. Extensive experiments with an unmanned ground vehicle in natural terrain demonstrate that the system adapts to complex environments with as little as 8 minutes of operational data, achieving a Matthews Correlation Coefficient (MCC) score of 0.63 and enabling safe navigation in densely vegetated environments. This work examines different training strategies for voxel-based TE methods and offers recommendations for training strategies to improve adaptability. The proposed method is validated on a robotic platform with limited computational resources (25W GPU), achieving accuracy comparable to offline-trained models while maintaining reliable performance across varied environments.

Evaluations of large language model (LLM) risks and capabilities are increasingly being incorporated into AI risk management and governance frameworks. Currently, most risk evaluations are conducted by designing inputs that elicit harmful behaviors from the system. However, a fundamental limitation of this approach is that the harmfulness of the behaviors identified during any particular evaluation can only lower bound the model's worst-possible-case behavior. As a complementary method for eliciting harmful behaviors, we propose evaluating LLMs with model tampering attacks which allow for modifications to latent activations or weights. We pit state-of-the-art techniques for removing harmful LLM capabilities against a suite of 5 input-space and 6 model tampering attacks. In addition to benchmarking these methods against each other, we show that (1) model resilience to capability elicitation attacks lies on a low-dimensional robustness subspace; (2) the attack success rate of model tampering attacks can empirically predict and offer conservative estimates for the success of held-out input-space attacks; and (3) state-of-the-art unlearning methods can easily be undone within 16 steps of fine-tuning. Together these results highlight the difficulty of removing harmful LLM capabilities and show that model tampering attacks enable substantially more rigorous evaluations than input-space attacks alone. We release models at https://huggingface.co/LLM-GAT

Inferring physical laws from data is a central challenge in science and engineering, including but not limited to healthcare, physical sciences, biosciences, social sciences, sustainability, climate, and robotics. Deep networks offer high-accuracy results but lack interpretability, prompting interest in models built from simple components. The Sparse Identification of Nonlinear Dynamics (SINDy) method has become the go-to approach for building such modular and interpretable models. SINDy leverages sparse regression with L1 regularization to identify key terms from a library of candidate functions. However, SINDy's choice of candidate library and optimization method requires significant technical expertise, limiting its widespread applicability. This work introduces Al-Khwarizmi, a novel agentic framework for physical law discovery from data, which integrates foundational models with SINDy. Leveraging LLMs, VLMs, and Retrieval-Augmented Generation (RAG), our approach automates physical law discovery, incorporating prior knowledge and iteratively refining candidate solutions via reflection. Al-Khwarizmi operates in two steps: it summarizes system observations-comprising textual descriptions, raw data, and plots-followed by a secondary step that generates candidate feature libraries and optimizer configurations to identify hidden physics laws correctly. Evaluating our algorithm on over 198 models, we demonstrate state-of-the-art performance compared to alternatives, reaching a 20 percent increase against the best-performing alternative.

Early identification of cognitive concerns is critical but often hindered by subtle symptom presentation. This study developed and validated a fully automated, multi-agent AI workflow using LLaMA 3 8B to identify cognitive concerns in 3,338 clinical notes from Mass General Brigham. The agentic workflow, leveraging task-specific agents that dynamically collaborate to extract meaningful insights from clinical notes, was compared to an expert-driven benchmark. Both workflows achieved high classification performance, with F1-scores of 0.90 and 0.91, respectively. The agentic workflow demonstrated improved specificity (1.00) and achieved prompt refinement in fewer iterations. Although both workflows showed reduced performance on validation data, the agentic workflow maintained perfect specificity. These findings highlight the potential of fully automated multi-agent AI workflows to achieve expert-level accuracy with greater efficiency, offering a scalable and cost-effective solution for detecting cognitive concerns in clinical settings.

Smishing is a social engineering attack using SMS containing malicious content to deceive individuals into disclosing sensitive information or transferring money to cybercriminals. Smishing attacks have surged by 328%, posing a major threat to mobile users, with losses exceeding \$54.2 million in 2019. Despite its growing prevalence, the issue remains significantly under-addressed. This paper presents a novel hybrid machine learning model for detecting Bangla smishing texts, combining Bidirectional Encoder Representations from Transformers (BERT) with Convolutional Neural Networks (CNNs) for enhanced character-level analysis. Our model addresses multi-class classification by distinguishing between Normal, Promotional, and Smishing SMS. Unlike traditional binary classification methods, our approach integrates BERT's contextual embeddings with CNN's character-level features, improving detection accuracy. Enhanced by an attention mechanism, the model effectively prioritizes crucial text segments. Our model achieves 98.47% accuracy, outperforming traditional classifiers, with high precision and recall in Smishing detection, and strong performance across all categories.

The robustness of Graph Neural Networks (GNNs) has become an increasingly important topic due to their expanding range of applications. Various attack methods have been proposed to explore the vulnerabilities of GNNs, ranging from Graph Modification Attacks (GMA) to the more practical and flexible Graph Injection Attacks (GIA). However, existing methods face two key challenges: (i) their reliance on surrogate models, which often leads to reduced attack effectiveness due to structural differences and prior biases, and (ii) existing GIA methods often sacrifice attack success rates in undefended settings to bypass certain defense models, thereby limiting their overall effectiveness. To overcome these limitations, we propose QUGIA, a Query-based and Unnoticeable Graph Injection Attack. QUGIA injects nodes by first selecting edges based on victim node connections and then generating node features using a Bayesian framework. This ensures that the injected nodes are similar to the original graph nodes, implicitly preserving homophily and making the attack more unnoticeable. Unlike previous methods, QUGIA does not rely on surrogate models, thereby avoiding performance degradation and achieving better generalization. Extensive experiments on six real-world datasets with diverse characteristics demonstrate that QUGIA achieves unnoticeable attacks and outperforms state-of-the-art attackers. The code will be released upon acceptance.

Language models exhibit complex, diverse behaviors when prompted with free-form text, making it difficult to characterize the space of possible outputs. We study the problem of behavior elicitation, where the goal is to search for prompts that induce specific target behaviors (e.g., hallucinations or harmful responses) from a target language model. To navigate the exponentially large space of possible prompts, we train investigator models to map randomly-chosen target behaviors to a diverse distribution of outputs that elicit them, similar to amortized Bayesian inference. We do this through supervised fine-tuning, reinforcement learning via DPO, and a novel Frank-Wolfe training objective to iteratively discover diverse prompting strategies. Our investigator models surface a variety of effective and human-interpretable prompts leading to jailbreaks, hallucinations, and open-ended aberrant behaviors, obtaining a 100% attack success rate on a subset of AdvBench (Harmful Behaviors) and an 85% hallucination rate.