To keep modern Radio Access Networks (RAN) running smoothly, operators need to spot the real-world triggers behind Service-Level Agreement (SLA) breaches well before customers feel them. We introduce an AI/ML pipeline that does two things most tools miss: (1) finds the likely root-cause indicators and (2) reveals the exact order in which those events unfold. We start by labeling network data: records linked to past SLA breaches are marked `abnormal', and everything else `normal'. Our model then learns the causal chain that turns normal behavior into a fault. In Monte Carlo tests the approach pinpoints the correct trigger sequence with high precision and scales to millions of data points without loss of speed. These results show that high-resolution, causally ordered insights can move fault management from reactive troubleshooting to proactive prevention.

Generative AI is reshaping offensive cybersecurity by enabling autonomous red team agents that can plan, execute, and adapt during penetration tests. However, existing approaches face trade-offs between generality and specialization, and practical deployments reveal challenges such as hallucinations, context limitations, and ethical concerns. In this work, we introduce a novel command & control (C2) architecture leveraging the Model Context Protocol (MCP) to coordinate distributed, adaptive reconnaissance agents covertly across networks. Notably, we find that our architecture not only improves goal-directed behavior of the system as whole, but also eliminates key host and network artifacts that can be used to detect and prevent command & control behavior altogether. We begin with a comprehensive review of state-of-the-art generative red teaming methods, from fine-tuned specialist models to modular or agentic frameworks, analyzing their automation capabilities against task-specific accuracy. We then detail how our MCP-based C2 can overcome current limitations by enabling asynchronous, parallel operations and real-time intelligence sharing without periodic beaconing. We furthermore explore advanced adversarial capabilities of this architecture, its detection-evasion techniques, and address dual-use ethical implications, proposing defensive measures and controlled evaluation in lab settings. Experimental comparisons with traditional C2 show drastic reductions in manual effort and detection footprint. We conclude with future directions for integrating autonomous exploitation, defensive LLM agents, predictive evasive maneuvers, and multi-agent swarms. The proposed MCP-enabled C2 framework demonstrates a significant step toward realistic, AI-driven red team operations that can simulate advanced persistent threats while informing the development of next-generation defensive systems.

Evaluating cross-lingual knowledge transfer in large language models is challenging, as correct answers in a target language may arise either from genuine transfer or from prior exposure during pre-training. We present LiveCLKTBench, an automated generation pipeline specifically designed to isolate and measure cross-lingual knowledge transfer. Our pipeline identifies self-contained, time-sensitive knowledge entities from real-world domains, filters them based on temporal occurrence, and verifies them against the model's knowledge. The documents of these valid entities are then used to generate factual questions, which are translated into multiple languages to evaluate transferability across linguistic boundaries. Using LiveCLKTBench, we evaluate several LLMs across five languages and observe that cross-lingual transfer is strongly influenced by linguistic distance and often asymmetric across language directions. While larger models improve transfer, the gains diminish with scale and vary across domains. These findings provide new insights into multilingual transfer and demonstrate the value of LiveCLKTBench as a reliable benchmark for future research.

Vision Transformers (ViTs) have demonstrated remarkable performance across a range of computer vision tasks; however, their high computational, memory, and energy demands hinder deployment on resource-constrained platforms. In this paper, we propose \emph{Cascaded-ViT (CViT)}, a lightweight and compute-efficient vision transformer architecture featuring a novel feedforward network design called \emph{Cascaded-Chunk Feed Forward Network (CCFFN)}. By splitting input features, CCFFN improves parameter and FLOP efficiency without sacrificing accuracy. Experiments on ImageNet-1K show that our \emph{CViT-XL} model achieves 75.5\% Top-1 accuracy while reducing FLOPs by 15\% and energy consumption by 3.3\% compared to EfficientViT-M5. Across various model sizes, the CViT family consistently exhibits the lowest energy consumption, making it suitable for deployment on battery-constrained devices such as mobile phones and drones. Furthermore, when evaluated using a new metric called \emph{Accuracy-Per-FLOP (APF)}, which quantifies compute efficiency relative to accuracy, CViT models consistently achieve top-ranking efficiency. Particularly, CViT-L is 2.2\% more accurate than EfficientViT-M2 while having comparable APF scores.

Modern language models fail a fundamental requirement of trustworthy intelligence: knowing when not to answer. Despite achieving impressive accuracy on benchmarks, these models produce confident hallucinations, even when wrong answers carry catastrophic consequences. Our evaluations on GSM8K, MedQA and GPQA show frontier models almost never abstain despite explicit warnings of severe penalties, suggesting that prompts cannot override training that rewards any answer over no answer. As a remedy, we propose Reinforced Hesitation (RH): a modification to Reinforcement Learning from Verifiable Rewards (RLVR) to use ternary rewards (+1 correct, 0 abstention, -$λ$ error) instead of binary. Controlled experiments on logic puzzles reveal that varying $λ$ produces distinct models along a Pareto frontier, where each training penalty yields the optimal model for its corresponding risk regime: low penalties produce aggressive answerers, high penalties conservative abstainers. We then introduce two inference strategies that exploit trained abstention as a coordination signal: cascading routes queries through models with decreasing risk tolerance, while self-cascading re-queries the same model on abstention. Both outperform majority voting with lower computational cost. These results establish abstention as a first-class training objective that transforms ``I don't know'' from failure into a coordination signal, enabling models to earn trust through calibrated honesty about their limits.

Both humans and machine learning models learn from experience, particularly in safety- and reliability-critical domains. While psychology seeks to understand human cognition, the field of Explainable AI (XAI) develops methods to interpret machine learning models. This study bridges these domains by applying computational tools from XAI to analyze human learning. We modeled human behavior during a complex real-world task -- tuning a particle accelerator -- by constructing graphs of operator subtasks. Applying techniques such as community detection and hierarchical clustering to archival operator data, we reveal how operators decompose the problem into simpler components and how these problem-solving structures evolve with expertise. Our findings illuminate how humans develop efficient strategies in the absence of globally optimal solutions, and demonstrate the utility of XAI-based methods for quantitatively studying human cognition.

Large Language Models (LLMs) excel at single-turn tasks such as instruction following and summarization, yet real-world deployments require sustained multi-turn interactions where user goals and conversational context persist and evolve. A recurring challenge in this setting is context drift: the gradual divergence of a model's outputs from goal-consistent behavior across turns. Unlike single-turn errors, drift unfolds temporally and is poorly captured by static evaluation metrics. In this work, we present a study of context drift in multi-turn interactions and propose a simple dynamical framework to interpret its behavior. We formalize drift as the turn-wise KL divergence between the token-level predictive distributions of the test model and a goal-consistent reference model, and propose a recurrence model that interprets its evolution as a bounded stochastic process with restoring forces and controllable interventions. We instantiate this framework in both synthetic long-horizon rewriting tasks and realistic user-agent simulations such as in $τ$-Bench, measuring drift for several open-weight LLMs that are used as user simulators. Our experiments consistently reveal stable, noise-limited equilibria rather than runaway degradation, and demonstrate that simple reminder interventions reliably reduce divergence in line with theoretical predictions. Together, these results suggest that multi-turn drift can be understood as a controllable equilibrium phenomenon rather than as inevitable decay, providing a foundation for studying and mitigating context drift in extended interactions.

Multimodal large language models (MLLMs) are increasingly used in real world applications, yet their safety under adversarial conditions remains underexplored. This study evaluates the harmlessness of four leading MLLMs (GPT-4o, Claude Sonnet 3.5, Pixtral 12B, and Qwen VL Plus) when exposed to adversarial prompts across text-only and multimodal formats. A team of 26 red teamers generated 726 prompts targeting three harm categories: illegal activity, disinformation, and unethical behaviour. These prompts were submitted to each model, and 17 annotators rated 2,904 model outputs for harmfulness using a 5-point scale. Results show significant differences in vulnerability across models and modalities. Pixtral 12B exhibited the highest rate of harmful responses (~62%), while Claude Sonnet 3.5 was the most resistant (~10%). Contrary to expectations, text-only prompts were slightly more effective at bypassing safety mechanisms than multimodal ones. Statistical analysis confirmed that both model type and input modality were significant predictors of harmfulness. These findings underscore the urgent need for robust, multimodal safety benchmarks as MLLMs are deployed more widely.

Access to reliable mental health information is vital for early help-seeking, yet expanding knowledge bases is resource-intensive and often misaligned with user needs. This results in poor performance of retrieval systems when presented concerns are not covered or expressed in informal or contextualized language. We present an AI-based gap-informed framework for corpus augmentation that authentically identifies underrepresented topics (gaps) by overlaying naturalistic user data such as forum posts in order to prioritize expansions based on coverage and usefulness. In a case study, we compare Directed (gap-informed augmentations) with Non-Directed augmentation (random additions), evaluating the relevance and usefulness of retrieved information across four retrieval-augmented generation (RAG) pipelines. Directed augmentation achieved near-optimal performance with modest expansions--requiring only a 42% increase for Query Transformation, 74% for Reranking and Hierarchical, and 318% for Baseline--to reach ~95% of the performance of an exhaustive reference corpus. In contrast, Non-Directed augmentation required substantially larger and thus practically infeasible expansions to achieve comparable performance (232%, 318%, 403%, and 763%, respectively). These results show that strategically targeted corpus growth can reduce content creation demands while sustaining high retrieval and provision quality, offering a scalable approach for building trusted health information repositories and supporting generative AI applications in high-stakes domains.

We introduce DeepFleet, a suite of foundation models designed to support coordination and planning for large-scale mobile robot fleets. These models are trained on fleet movement data, including robot positions, goals, and interactions, from hundreds of thousands of robots in Amazon warehouses worldwide. DeepFleet consists of four architectures that each embody a distinct inductive bias and collectively explore key points in the design space for multi-agent foundation models: the robot-centric (RC) model is an autoregressive decision transformer operating on neighborhoods of individual robots; the robot-floor (RF) model uses a transformer with cross-attention between robots and the warehouse floor; the image-floor (IF) model applies convolutional encoding to a multi-channel image representation of the full fleet; and the graph-floor (GF) model combines temporal attention with graph neural networks for spatial relationships. In this paper, we describe these models and present our evaluation of the impact of these design choices on prediction task performance. We find that the robot-centric and graph-floor models, which both use asynchronous robot state updates and incorporate the localized structure of robot interactions, show the most promise. We also present experiments that show that these two models can make effective use of larger warehouses operation datasets as the models are scaled up.