PAC generalization bounds on the risk, when expressed in terms of the expected loss, are often insufficient to capture imbalances between subgroups in the data. To overcome this limitation, we introduce a new family of risk measures, called constrained f-entropic risk measures, which enable finer control over distributional shifts and subgroup imbalances via f-divergences, and include the Conditional Value at Risk (CVaR), a well-known risk measure. We derive both classical and disintegrated PAC-Bayesian generalization bounds for this family of risks, providing the first disintegratedPAC-Bayesian guarantees beyond standard risks. Building on this theory, we design a self-bounding algorithm that minimizes our bounds directly, yielding models with guarantees at the subgroup level. Finally, we empirically demonstrate the usefulness of our approach.

We propose MAE-SAM2, a novel foundation model for retinal vascular leakage segmentation on fluorescein angiography images. Due to the small size and dense distribution of the leakage areas, along with the limited availability of labeled clinical data, this presents a significant challenge for segmentation tasks. Our approach integrates a Self-Supervised learning (SSL) strategy, Masked Autoencoder (MAE), with SAM2. In our implementation, we explore different loss functions and conclude a task-specific combined loss. Extensive experiments and ablation studies demonstrate that MAE-SAM2 outperforms several state-of-the-art models, achieving the highest Dice score and Intersection-over-Union (IoU). Compared to the original SAM2, our model achieves a $5\%$ performance improvement, highlighting the promise of foundation models with self-supervised pretraining in clinical imaging tasks.

Satisfiability Modulo Theory (SMT) solvers are foundational to modern systems and programming languages research, providing the foundation for tasks like symbolic execution and automated verification. Because these solvers sit on the critical path, their correctness is essential, and high-quality test formulas are key to uncovering bugs. However, while prior testing techniques performed well on earlier solver versions, they struggle to keep pace with rapidly evolving features. Recent approaches based on Large Language Models (LLMs) show promise in exploring advanced solver capabilities, but two obstacles remain: nearly half of the generated formulas are syntactically invalid, and iterative interactions with LLMs introduce substantial computational overhead. In this study, we present Once4All, a novel LLM-assisted fuzzing framework that addresses both issues by shifting from direct formula generation to the synthesis of generators for reusable terms (i.e., logical expressions). Specifically, Once4All uses LLMs to (1) automatically extract context-free grammars (CFGs) for SMT theories, including solver-specific extensions, from documentation, and (2) synthesize composable Boolean term generators that adhere to these grammars. During fuzzing, Once4All populates structural skeletons derived from existing formulas with the terms iteratively produced by the LLM-synthesized generators. This design ensures syntactic validity while promoting semantic diversity. Notably, Once4All requires only one-time LLM interaction investment, dramatically reducing runtime cost. We evaluated Once4All on two leading SMT solvers: Z3 and cvc5. Our experiments show that Once4All has identified 43 confirmed bugs, 40 of which have already been fixed by developers.

On-device running Large Language Models (LLMs) is nowadays a critical enabler towards preserving user privacy. We observe that the attention operator falls back from the special-purpose NPU to the general-purpose CPU/GPU because of quantization sensitivity in state-of-the-art frameworks. This fallback results in a degraded user experience and increased complexity in system scheduling. To this end, this paper presents shadowAttn, a system-algorithm codesigned sparse attention module with minimal reliance on CPU/GPU by only sparsely calculating the attention on a tiny portion of tokens. The key idea is to hide the overhead of estimating the important tokens with a NPU-based pilot compute. Further, shadowAttn proposes insightful techniques such as NPU compute graph bucketing, head-wise NPU-CPU/GPU pipeline and per-head fine-grained sparsity ratio to achieve high accuracy and efficiency. shadowAttn delivers the best performance with highly limited CPU/GPU resource; it requires much less CPU/GPU resource to deliver on-par performance of SoTA frameworks.

The increasing sophistication of large vision-language models (LVLMs) has been accompanied by advances in safety alignment mechanisms designed to prevent harmful content generation. However, these defenses remain vulnerable to sophisticated adversarial attacks. Existing jailbreak methods typically rely on direct and semantically explicit prompts, overlooking subtle vulnerabilities in how LVLMs compose information over multiple reasoning steps. In this paper, we propose a novel and effective jailbreak framework inspired by Return-Oriented Programming (ROP) techniques from software security. Our approach decomposes a harmful instruction into a sequence of individually benign visual gadgets. A carefully engineered textual prompt directs the sequence of inputs, prompting the model to integrate the benign visual gadgets through its reasoning process to produce a coherent and harmful output. This makes the malicious intent emergent and difficult to detect from any single component. We validate our method through extensive experiments on established benchmarks including SafeBench and MM-SafetyBench, targeting popular LVLMs. Results show that our approach consistently and substantially outperforms existing baselines on state-of-the-art models, achieving near-perfect attack success rates (over 0.90 on SafeBench) and improving ASR by up to 0.39. Our findings reveal a critical and underexplored vulnerability that exploits the compositional reasoning abilities of LVLMs, highlighting the urgent need for defenses that secure the entire reasoning process.

Due to limited computational resources, medium-range temperature forecasts typically rely on low-resolution numerical weather prediction (NWP) models, which are prone to systematic and random errors. We propose a method that integrates a convolutional neural network (CNN) with an ensemble of low-resolution NWP models (40-km horizontal resolution) to produce high-resolution (5-km) surface temperature forecasts with lead times extending up to 5.5 days (132 h). First, CNN-based post-processing (bias correction and spatial downscaling) is applied to individual ensemble members to reduce systematic errors and perform downscaling, which improves the deterministic forecast accuracy. Second, this member-wise correction is applied to all 51 ensemble members to construct a new high-resolution ensemble forecasting system with an improved probabilistic reliability and spread-skill ratio that differs from the simple error reduction mechanism of ensemble averaging. Whereas averaging reduces forecast errors by smoothing spatial fields, our member-wise CNN correction reduces error from noise while maintaining forecast information at a level comparable to that of other high-resolution forecasts. Experimental results indicate that the proposed method provides a practical and scalable solution for improving medium-range temperature forecasts, which is particularly valuable for use in operational centers with limited computational resources.

Stochastic simulators exhibit intrinsic stochasticity due to unobservable, uncontrollable, or unmodeled input variables, resulting in random outputs even at fixed input conditions. Such simulators are common across various scientific disciplines; however, emulating their entire conditional probability distribution is challenging, as it is a task traditional deterministic surrogate modeling techniques are not designed for. Additionally, accurately characterizing the response distribution can require prohibitively large datasets, especially for computationally expensive high-fidelity (HF) simulators. When lower-fidelity (LF) stochastic simulators are available, they can enhance limited HF information within a multifidelity surrogate modeling (MFSM) framework. While MFSM techniques are well-established for deterministic settings, constructing multifidelity emulators to predict the full conditional response distribution of stochastic simulators remains a challenge. In this paper, we propose multifidelity generalized lambda models (MF-GLaMs) to efficiently emulate the conditional response distribution of HF stochastic simulators by exploiting data from LF stochastic simulators. Our approach builds upon the generalized lambda model (GLaM), which represents the conditional distribution at each input by a flexible, four-parameter generalized lambda distribution. MF-GLaMs are non-intrusive, requiring no access to the internal stochasticity of the simulators nor multiple replications of the same input values. We demonstrate the efficacy of MF-GLaM through synthetic examples of increasing complexity and a realistic earthquake application. Results show that MF-GLaMs can achieve improved accuracy at the same cost as single-fidelity GLaMs, or comparable performance at significantly reduced cost.

Embodied AI systems, comprising AI models and physical plants, are increasingly prevalent across various applications. Due to the rarity of system failures, ensuring their safety in complex operating environments remains a major challenge, which severely hinders their large-scale deployment in safety-critical domains, such as autonomous vehicles, medical devices, and robotics. While achieving provable deterministic safety-verifying system safety across all possible scenarios-remains theoretically ideal, the rarity and complexity of corner cases make this approach impractical for scalable embodied AI systems. Instead, empirical safety evaluation is employed as an alternative, but the absence of provable guarantees imposes significant limitations. To address these issues, we argue for a paradigm shift to provable probabilistic safety that integrates provable guarantees with progressive achievement toward a probabilistic safety boundary on overall system performance. The new paradigm better leverages statistical methods to enhance feasibility and scalability, and a well-defined probabilistic safety boundary enables embodied AI systems to be deployed at scale. In this Perspective, we outline a roadmap for provable probabilistic safety, along with corresponding challenges and potential solutions. By bridging the gap between theoretical safety assurance and practical deployment, this Perspective offers a pathway toward safer, large-scale adoption of embodied AI systems in safety-critical applications.

Large Language Models (LLMs) have substantially improved the conversational capabilities of social robots. Nevertheless, for an intuitive and fluent human-robot interaction, robots should be able to ground the conversation by relating ambiguous or underspecified spoken utterances to the current physical situation and to the intents expressed nonverbally by the user, such as through referential gaze. Here, we propose a representation that integrates speech and gaze to enable LLMs to achieve higher situated awareness and correctly resolve ambiguous requests. Our approach relies on a text-based semantic translation of the scanpath produced by the user, along with the verbal requests. It demonstrates LLMs' capabilities to reason about gaze behavior, robustly ignoring spurious glances or irrelevant objects. We validate the system across multiple tasks and two scenarios, showing its superior generality and accuracy compared to control conditions. We demonstrate an implementation on a robotic platform, closing the loop from request interpretation to execution.

In both machine learning and in computational neuroscience, plasticity in functional neural networks is frequently expressed as gradient descent on a cost. Often, this imposes symmetry constraints that are difficult to reconcile with local computation, as is required for biological networks or neuromorphic hardware. For example, wake-sleep learning in networks characterized by Boltzmann distributions assumes symmetric connectivity. Similarly, the error backpropagation algorithm is notoriously plagued by the weight transport problem between the representation and the error stream. Existing solutions such as feedback alignment circumvent the problem by deferring to the robustness of these algorithms to weight asymmetry. However, they scale poorly with network size and depth. We introduce spike-based alignment learning (SAL), a complementary learning rule for spiking neural networks, which uses spike timing statistics to extract and correct the asymmetry between effective reciprocal connections. Apart from being spike-based and fully local, our proposed mechanism takes advantage of noise. Based on an interplay between Hebbian and anti-Hebbian plasticity, synapses can thereby recover the true local gradient. This also alleviates discrepancies that arise from neuron and synapse variability -- an omnipresent property of physical neuronal networks. We demonstrate the efficacy of our mechanism using different spiking network models. First, SAL can significantly improve convergence to the target distribution in probabilistic spiking networks versus Hebbian plasticity alone. Second, in neuronal hierarchies based on cortical microcircuits, SAL effectively aligns feedback weights to the forward pathway, thus allowing the backpropagation of correct feedback errors. Third, our approach enables competitive performance in deep networks using only local plasticity for weight transport.

Two-time-scale stochastic approximation algorithms are iterative methods used in applications such as optimization, reinforcement learning, and control. Finite-time analysis of these algorithms has primarily focused on fixed point iterations where both time-scales have contractive mappings. In this work, we broaden the scope of such analyses by considering settings where the slower time-scale has a non-expansive mapping. For such algorithms, the slower time-scale can be viewed as a stochastic inexact Krasnoselskii-Mann iteration. We also study a variant where the faster time-scale has a projection step which leads to non-expansiveness in the slower time-scale. We show that the last-iterate mean square residual error for such algorithms decays at a rate $O(1/k^{1/4-ε})$, where $ε>0$ is arbitrarily small. We further establish almost sure convergence of iterates to the set of fixed points. We demonstrate the applicability of our framework by applying our results to minimax optimization, linear stochastic approximation, and Lagrangian optimization.

Best Arm Identification (BAI) problems are progressively used for data-sensitive applications, such as designing adaptive clinical trials, tuning hyper-parameters, and conducting user studies. Motivated by the data privacy concerns invoked by these applications, we study the problem of BAI with fixed confidence in both the local and central models, i.e. $ε$-local and $ε$-global Differential Privacy (DP). First, to quantify the cost of privacy, we derive lower bounds on the sample complexity of any $δ$-correct BAI algorithm satisfying $ε$-global DP or $ε$-local DP. Our lower bounds suggest the existence of two privacy regimes. In the high-privacy regime, the hardness depends on a coupled effect of privacy and novel information-theoretic quantities involving the Total Variation. In the low-privacy regime, the lower bounds reduce to the non-private lower bounds. We propose $ε$-local DP and $ε$-global DP variants of a Top Two algorithm, namely CTB-TT and AdaP-TT*, respectively. For $ε$-local DP, CTB-TT is asymptotically optimal by plugging in a private estimator of the means based on Randomised Response. For $ε$-global DP, our private estimator of the mean runs in arm-dependent adaptive episodes and adds Laplace noise to ensure a good privacy-utility trade-off. By adapting the transportation costs, the expected sample complexity of AdaP-TT* reaches the asymptotic lower bound up to multiplicative constants.

This paper develops a viable notion of learning for sampling-based algorithms that applies in broader settings than previously considered. More specifically, we model a discounted infinite-horizon MDPs with Borel state and action spaces, whose rewards and transitions depend on an unknown parameter. To analyze adaptive learning algorithms based on sampling we introduce a general canonical probability space in this setting. Since standard definitions of regret are inadequate for policy evaluation in this setting, we propose new metrics that arise from decomposing the standard expected regret in discounted infinite-horizon MDPs into three terms: (i) the expected finite-time regret, (ii) the expected state regret, and (iii) the expected residual regret. Component (i) translates into the traditional concept of expected regret over a finite horizon. Term (ii) reflects how much future performance is compromised at a given time because earlier decisions have led the system to a less favorable state than under an optimal policy. Finally, metric (iii) measures regret with respect to the optimal reward from the current period onward, disregarding the irreversible consequences of past decisions. We further disaggregate this term by introducing the probabilistic residual regret, a finer, sample-path version of (iii) that captures the remaining loss in future performance from the current period onward, conditional on the observed history. Its expectation coincides with (iii). We then focus on Thompson sampling (TS); under assumptions that extend those used in prior work on finite state and action spaces to the Borel setting, we show that component (iii) for TS converges to zero exponentially fast. We further show that, under mild conditions ensuring the existence of the relevant limits, its probabilistic counterpart converges to zero almost surely and TS achieves complete learning.

We study the linearization of a discrete transportation distance between probability distributions on finite weighted graphs originally due to Maas (``Gradient flows of the entropy for finite {M}arkov chains,'' J. Funct. Anal. 261(8), 2011) which demonstrates various connections to the underlying combinatorial structure of the graph. For a connected graph and a reference density $μ$ on its vertices, our main result is a nonasymptotic local linearization theorem showing that if $ν$ is a small additive perturbation of $μ$ then their squared discrete transportation distance is controlled above and below by the quadratic form of the pseudoinverse of a re-weighted graph Laplacian matrix. When the reference measure is stationary for the simple random walk on the graph, the weights agree with the original graph and this yields the quadratic form $(μ-ν)^\top L_w^\dagger (μ-ν)$, which can be viewed as a form of resistance distance between probability measures. This distance has a number of combinatorial and variational characterizations, including Beckmann and Benamou--Brenier formulas, a dual homogeneous Sobolev norm formula, a spanning $2$-forest formula, and a representation through random walk hitting times. Finally, we show that on the resulting ``resistance manifold,'' the gradient flow of the $χ^2$ functional is the continuous-time random walk and that its geodesic strong convexity modulus equals the spectral gap of the normalized Laplacian. From this geometric vantage point, one recovers the classical fact that the spectral gap of the normalized Laplacian controls the exponential convergence rate of the random walk to stationarity.

Inspired by the concept of active learning, we propose active inference$\unicode{x2013}$a methodology for statistical inference with machine-learning-assisted data collection. Assuming a budget on the number of labels that can be collected, the methodology uses a machine learning model to identify which data points would be most beneficial to label, thus effectively utilizing the budget. It operates on a simple yet powerful intuition: prioritize the collection of labels for data points where the model exhibits uncertainty, and rely on the model's predictions where it is confident. Active inference constructs provably valid confidence intervals and hypothesis tests while leveraging any black-box machine learning model and handling any data distribution. The key point is that it achieves the same level of accuracy with far fewer samples than existing baselines relying on non-adaptively-collected data. This means that for the same number of collected samples, active inference enables smaller confidence intervals and more powerful p-values. We evaluate active inference on datasets from public opinion research, census analysis, and proteomics.

We develop novel learning rates for conditional mean embeddings by applying the theory of interpolation for reproducing kernel Hilbert spaces (RKHS). We derive explicit, adaptive convergence rates for the sample estimator under the misspecifed setting, where the target operator is not Hilbert-Schmidt or bounded with respect to the input/output RKHSs. We demonstrate that in certain parameter regimes, we can achieve uniform convergence rates in the output RKHS. We hope our analyses will allow the much broader application of conditional mean embeddings to more complex ML/RL settings involving infinite dimensional RKHSs and continuous state spaces.

The rapid growth of generative artificial intelligence (AI) has introduced unprecedented computational demands, driving significant increases in the energy footprint of data centers. However, existing power consumption data is largely proprietary and reported at varying resolutions, creating challenges for estimating whole-facility energy use and planning infrastructure. In this work, we present a methodology that bridges this gap by linking high-resolution workload power measurements to whole-facility energy demand. Using NLR's high-performance computing data center equipped with NVIDIA H100 GPUs, we measure power consumption of AI workloads at 0.1-second resolution for AI training, fine-tuning and inference jobs. Workloads are characterized using MLCommons benchmarks for model training and fine-tuning, and vLLM benchmarks for inference, enabling reproducible and standardized workload profiling. The dataset of power consumption profiles is made publicly available. These power profiles are then scaled to the whole-facility-level using a bottom-up, event-driven, data center energy model. The resulting whole-facility energy profiles capture realistic temporal fluctuations driven by AI workloads and user-behavior, and can be used to inform infrastructure planning for grid connection, on-site energy generation, and distributed microgrids.

In this paper, we derive rates of convergence in the high-dimensional central limit theorem for Polyak-Ruppert averaged iterates generated by the asynchronous Q-learning algorithm with a polynomial stepsize $k^{-ω},\, ω\in (1/2, 1]$. Assuming that the sequence of state-action-next-state triples $(s_k, a_k, s_{k+1})_{k \geq 0}$ forms a uniformly geometrically ergodic Markov chain, we establish a rate of order up to $n^{-1/6} \log^{4} (nS A)$ over the class of hyper-rectangles, where $n$ is the number of samples used by the algorithm and $S$ and $A$ denote the numbers of states and actions, respectively. To obtain this result, we prove a high-dimensional central limit theorem for sums of martingale differences, which may be of independent interest. Finally, we present bounds for high-order moments for the algorithm's last iterate.

Propositional Linear Temporal Logic (LTL) is a popular formalism for specifying desirable requirements and security and privacy policies for software, networks, and systems. Yet expressing such requirements and policies in LTL remains challenging because of its intricate semantics. Since many security and privacy analysis tools require LTL formulas as input, this difficulty places them out of reach for many developers and analysts. Large Language Models (LLMs) could broaden access to such tools by translating natural language fragments into LTL formulas. This paper evaluates that premise by assessing how effectively several representative LLMs translate assertive English sentences into LTL formulas. Using both human-generated and synthetic ground-truth data, we evaluate effectiveness along syntactic and semantic dimensions. The results reveal three findings: (1) in line with prior findings, LLMs perform better on syntactic aspects of LTL than on semantic ones; (2) they generally benefit from more detailed prompts; and (3) reformulating the task as a Python code-completion problem substantially improves overall performance. We also discuss challenges in conducting a fair evaluation on this task and conclude with recommendations for future work.

Large Language Models (LLMs) challenge conventional automated programming assessment because students can now produce functionally correct code without demonstrating corresponding understanding. This paper makes two contributions. First, it reports a saturation-based scoping review of conversational assessment approaches in programming education. The review identifies three dominant architectural families: rule-based or template-driven systems, LLM-based systems, and hybrid systems. Across the literature, conversational agents appear promising for scalable feedback and deeper probing of code understanding, but important limitations remain around hallucinations, over-reliance, privacy, integrity, and deployment constraints. Second, the paper synthesizes these findings into a Hybrid Socratic Framework for integrating conversational verification into Automated Programming Assessment Systems (APASs). The framework combines deterministic code analysis with a dual-agent conversational layer, knowledge tracking, scaffolded questioning, and guardrails that tie prompts to runtime facts. The paper also discusses practical safeguards against LLM-generated explanations, including proctored deployment modes, randomized trace questions, stepwise reasoning tied to concrete execution states, and local-model deployment options for privacy-sensitive settings. Rather than replacing conventional testing, the framework is intended as a complementary layer for verifying whether students understand the code they submit.

GROMACS is a de-facto standard for classical Molecular Dynamics (MD). The rise of AI-driven interatomic potentials that pursue near-quantum accuracy at MD throughput now poses a significant challenge: embedding neural-network inference into multi-GPU simulations retaining high-performance. In this work, we integrate the MLIP framework DeePMD-kit into GROMACS, enabling domain-decomposed, GPU-accelerated inference across multi-node systems. We extend the GROMACS NNPot interface with a DeePMD backend, and we introduce a domain decomposition layer decoupled from the main simulation. The inference is executed concurrently on all processes, with two MPI collectives used each step to broadcast coordinates and to aggregate and redistribute forces. We train an in-house DPA-1 model (1.6 M parameters) on a dataset of solvated protein fragments. We validate the implementation on a small protein system, then we benchmark the GROMACS-DeePMD integration with a 15,668 atom protein on NVIDIA A100 and AMD MI250x GPUs up to 32 devices. Strong-scaling efficiency reaches 66% at 16 devices and 40% at 32; weak-scaling efficiency is 80% to 16 devices and reaches 48% (MI250x) and 40% (A100) at 32 devices. Profiling with the ROCm System profiler shows that >90% of the wall time is spent in DeePMD inference, while MPI collectives contribute <10%, primarily since they act as a global synchronization point. The principal bottlenecks are the irreducible ghost-atom cost set by the cutoff radius, confirmed by a simple throughput model, and load imbalance across ranks. These results demonstrate that production MD with near ab initio fidelity is feasible at scale in GROMACS.

Gaussian process ($GP$) regression is a widely used non-parametric modeling tool, but its cubic complexity in the training size limits its use on massive data sets. A practical remedy is to predict using only the nearest neighbours of each test point, as in Nearest Neighbour Gaussian Process ($NNGP$) regression for geospatial problems and the related scalable $GPnn$ method for more general machine-learning applications. Despite their strong empirical performance, the large-$n$ theory of $NNGP/GPnn$ remains incomplete. We develop a theoretical framework for $NNGP$ and $GPnn$ regression. Under mild regularity assumptions, we derive almost sure pointwise limits for three key predictive criteria: mean squared error ($MSE$), calibration coefficient ($CAL$), and negative log-likelihood ($NLL$). We then study the $L_2$-risk, prove universal consistency, and show that the risk attains Stone's minimax rate $n^{-2α/(2p+d)}$, where $α$ and $p$ capture regularity of the regression problem. We also prove uniform convergence of $MSE$ over compact hyper-parameter sets and show that its derivatives with respect to lengthscale, kernel scale, and noise variance vanish asymptotically, with explicit rates. This explains the observed robustness of $GPnn$ to hyper-parameter tuning. These results provide a rigorous statistical foundation for $NNGP/GPnn$ as a highly scalable and principled alternative to full $GP$ models.

Operating LEO mega-constellations requires translating high-level operator intents ("reroute financial traffic away from polar links under 80 ms") into low-level routing constraints -- a task that demands both natural language understanding and network-domain expertise. We present an end-to-end system comprising three components: (1) a GNN cost-to-go router that distills Dijkstra-quality routing into a 152K-parameter graph attention network achieving 99.8% packet delivery ratio with 17x inference speedup; (2) an LLM intent compiler that converts natural language to a typed constraint intermediate representation using few-shot prompting with a verifier-feedback repair loop, achieving 98.4% compilation rate and 87.6% full semantic match on feasible intents in a 240-intent benchmark (193 feasible, 47 infeasible); and (3) an 8-pass deterministic validator with constructive feasibility certification that achieves 0% unsafe acceptance on all 47 infeasible intents (30 labeled + 17 discovered by Pass 8), with 100% corruption detection across 240 structural corruption tests and 100% on 15 targeted adversarial attacks. End-to-end evaluation across four constrained routing scenarios confirms zero constraint violations with both routers. We further demonstrate that apparent performance gaps in polar-avoidance scenarios are largely explained by topological reachability ceilings rather than routing quality, and that the LLM compiler outperforms a rule-based baseline by 46.2 percentage points on compositional intents. Our system bridges the semantic gap between operator intent and network configuration while maintaining the safety guarantees required for operational deployment.

Existing driving automation (DA) systems on production vehicles rely on human drivers to decide when to engage DA while requiring them to remain continuously attentive and ready to intervene. This design demands substantial situational judgment and imposes significant cognitive load, leading to steep learning curves, suboptimal user experience, and safety risks from both over-reliance and delayed takeover. Predicting when drivers hand over control to DA and when they take it back is therefore critical for designing proactive, context-aware HMI, yet existing datasets rarely capture the multimodal context, including road scene, driver state, vehicle dynamics, and route environment. To fill this gap, we introduce BATON, a large-scale naturalistic dataset capturing real-world DA usage across 127 drivers, and 136.6 hours of driving. The dataset synchronizes front-view video, in-cabin video, decoded CAN bus signals, radar-based lead-vehicle interaction, and GPS-derived route context, forming a closed-loop multimodal record around each control transition. We define three benchmark tasks: driving action understanding, handover prediction, and takeover prediction, and evaluate baselines spanning sequence models, classical classifiers, and zero-shot VLMs. Results show that visual input alone is insufficient for reliable transition prediction: front-view video captures road context but not driver state, while in-cabin video reflects driver readiness but not the external scene. Incorporating CAN and route-context signals substantially improves performance over video-only settings, indicating strong complementarity across modalities. We further find takeover events develop more gradually and benefit from longer prediction horizons, whereas handover events depend more on immediate contextual cues, revealing an asymmetry with direct implications for HMI design in assisted driving systems.

In gender-restrictive and surveilled contexts, where access to formal education may be restricted for women, pursuing education involves safety and privacy risks. When women are excluded from schools and universities, they often turn to online self-learning and generative AI (GenAI) to pursue their educational and career aspirations. However, we know little about what safe and accountable GenAI support is required in the context of surveillance, household responsibilities, and the absence of learning communities. We present a remote participatory design study with 20 women in Afghanistan, informed by a recruitment survey (n = 140), examining how participants envision GenAI for learning and employability. Participants describe using GenAI less as an information source and more as an always-available peer, mentor, and source of career guidance that helps compensate for the absence of learning communities. At the same time, they emphasize that this companionship is constrained by privacy and surveillance risks, contextually unrealistic and culturally unsafe support, and direct-answer interactions that can undermine learning by creating an illusion of progress. Beyond eliciting requirements, envisioning the future with GenAI through participatory design was positively associated with significant increases in participants' aspirations (p=.01), perceived agency (p=.01), and perceived avenues (p=.03). These outcomes show that accountable and safe GenAI is not only about harm reduction but can also actively enable women to imagine and pursue viable learning and employment futures. Building on this, we translate participants' proposals into accountability-focused design directions that center on safety-first interaction and user control, context-grounded support under constrained resources, and offer pedagogically aligned assistance that supports genuine learning rather than quick answers.

We introduce a code-based challenge for automated, open-ended mathematical discovery based on the $k$-server conjecture, a central open problem in competitive analysis. The task is to discover a potential function satisfying a large graph-structured system of simple linear inequalities. The resulting evaluation procedure is sound but incomplete: any violated inequality definitively refutes a candidate, whereas satisfying all inequalities does not by itself constitute a proof of the corresponding conjecture's special case. Nevertheless, a candidate that passes all constraints would be strong evidence toward a valid proof and, to the best of our knowledge, no currently known potential achieves this under our formulation in the open $k=4$ circle case. As such, a successful candidate would already be an interesting contribution to the $k$-server conjecture, and could become a substantial theoretical result when paired with a full proof. Experiments on the resolved $k=3$ regime show that current agentic methods can solve nontrivial instances, and in the open $k=4$ regime they reduce the number of violations relative to existing potentials without fully resolving the task. Taken together, these results suggest that the task is challenging but plausibly within reach of current methods. Beyond its relevance to the $k$-server community, where the developed tooling enables researchers to test new hypotheses and potentially improve on the current record, the task also serves as a useful \emph{benchmark} for developing code-based discovery agents. In particular, our $k=3$ results show that it mitigates important limitations of existing open-ended code-based benchmarks, including early saturation and the weak separation between naive random baselines and more sophisticated methods.

As large language models (LLMs) evolve from static chatbots into autonomous agents, the primary vulnerability surface shifts from final outputs to intermediate execution traces. While safety guardrails are well-benchmarked for natural language responses, their efficacy remains largely unexplored within multi-step tool-use trajectories. To address this gap, we introduce TraceSafe-Bench, the first comprehensive benchmark specifically designed to assess mid-trajectory safety. It encompasses 12 risk categories, ranging from security threats (e.g., prompt injection, privacy leaks) to operational failures (e.g., hallucinations, interface inconsistencies), featuring over 1,000 unique execution instances. Our evaluation of 13 LLM-as-a-guard models and 7 specialized guardrails yields three critical findings: 1) Structural Bottleneck: Guardrail efficacy is driven more by structural data competence (e.g., JSON parsing) than semantic safety alignment. Performance correlates strongly with structured-to-text benchmarks ($ρ=0.79$) but shows near-zero correlation with standard jailbreak robustness. 2) Architecture over Scale: Model architecture influences risk detection performance more significantly than model size, with general-purpose LLMs consistently outperforming specialized safety guardrails in trajectory analysis. 3) Temporal Stability: Accuracy remains resilient across extended trajectories. Increased execution steps allow models to pivot from static tool definitions to dynamic execution behaviors, actually improving risk detection performance in later stages. Our findings suggest that securing agentic workflows requires jointly optimizing for structural reasoning and safety alignment to effectively mitigate mid-trajectory risks.

Multimodal retrieval systems struggle to resolve image-text queries against text-only corpora: the best vision-language encoder achieves only 27.6 nDCG@10 on MM-BRIGHT, underperforming strong text-only retrievers. We argue the bottleneck is not the retriever but the query -- raw multimodal queries entangle visual descriptions, conversational noise, and retrieval intent in ways that systematically degrade embedding similarity. We present \textbf{BRIDGE}, a two-component system that resolves this mismatch without multimodal encoders. \textbf{FORGE} (\textbf{F}ocused Retrieval Query Generato\textbf{r}) is a query alignment model trained via reinforcement learning, which distills noisy multimodal queries into compact, retrieval-optimized search strings. \textbf{LENS} (\textbf{L}anguage-\textbf{E}nhanced \textbf{N}eural \textbf{S}earch) is a reasoning-enhanced dense retriever fine-tuned on reasoning-intensive retrieval data to handle the intent-rich queries FORGE produces. Evaluated on MM-BRIGHT (2,803 queries, 29 domains), BRIDGE achieves \textbf{29.7} nDCG@10, surpassing all multimodal encoder baselines including Nomic-Vision (27.6). When FORGE is applied as a plug-and-play aligner on top of Nomic-Vision, the combined system reaches \textbf{33.3} nDCG@10 -- exceeding the best text-only retriever (32.2) -- demonstrating that \textit{query alignment} is the key bottleneck in multimodal-to-text retrieval. https://github.com/mm-bright/multimodal-reasoning-retrieval

We investigate the feasibility of real-time 3D Gaussian Splatting (3DGS) rasterisation on edge clients with varying Gaussian splat counts and GPU computational budgets. Instead of evaluating multiple physical devices, we adopt an emulation-based approach that approximates different GPU capability tiers on a single high-end GPU. By systematically under-clocking the GPU core frequency and applying power caps, we emulate a controlled range of floating-point performance levels that approximate different GPU capability tiers. At each point in this range, we measure frame rate, runtime behaviour, and power consumption across scenes of varying complexity, pipelines, and optimisations, enabling analysis of power-performance relationships such as FPS-power curves, energy per frame, and performance per watt. This method allows us to approximate the performance envelope of a diverse class of GPUs, from embedded and mobile-class devices to high-end consumer-grade systems. Our objective is to explore the practical lower bounds of client-side 3DGS rasterisation and assess its potential for deployment in energy-constrained environments, including standalone headsets and thin clients. Through this analysis, we provide early insights into the performance-energy trade-offs that govern the viability of edge-deployed 3DGS systems.

This paper focuses on energy savings in downlink operation of cell-free massive MIMO (CF mMIMO) networks under dynamic traffic conditions. We propose a multi-agent deep reinforcement learning (MADRL) algorithm that enables each access point (AP) to autonomously control antenna re-configuration and advanced sleep mode (ASM) selection. After the training process, the proposed framework operates in a fully distributed manner, eliminating the need for centralized control and allowing each AP to dynamically adjust to real-time traffic fluctuations. Simulation results show that the proposed algorithm reduces power consumption (PC) by 56.23% compared to systems without any energy-saving scheme and by 30.12% relative to a non-learning mechanism that only utilizes the lightest sleep mode, with only a slight increase in drop ratio. Moreover, compared to the widely used deep Q-network (DQN) algorithm, it achieves a similar PC level but with a significantly lower drop ratio.