Computer Science & Mathematics

Dynamic metasurface antennas (DMAs) have emerged as a promising technology for realizing large-scale antenna arrays in next-generation wireless networks owing to their ability to enable dense and energy-efficient antenna architectures. However, jointly optimizing digital beamforming and DMA element weights under time-varying channel conditions remains challenging due to their coupled effects on the received signal and interference powers. We formulate this problem to maximize time-averaged energy efficiency (EE) under time-varying channel conditions, subject to transmit power constraints, Lorentzian phase limitations, and minimum rate requirements. To enable long-term planning and improve sample efficiency compared with model-free reinforcement learning, we develop a world model framework that learns latent system dynamics from interactions with the environment and jointly optimizes digital beamforming and DMA element weights through imagined trajectories. Furthermore, we incorporate feasibility-preserving action mappings and safety cost signals to ensure constraint satisfaction during policy learning. Simulation results demonstrate that the proposed framework improves the time-averaged EE by approximately 20% compared with the deep Q-network baseline, while achieving higher sample efficiency and more reliable constraint satisfaction.

The security region (SR) specifies a joint feasible region in the operational space involving various installations within the power system. It enables dynamic allocation of capacity among these energy units, thereby unlocking additional grid potential for power delivery and mitigating congestion. By constructing a convex hull from a set of vertices to form a polytope, the SR can be effectively represented with an acceptable accuracy loss. However, the SR computation inevitably generates many low-quality vertices, which contribute little to the overall hypervolume and can be discarded. By introducing SR compression, the SR data size can be reduced, which in turn facilitates the memory-efficient computation of SR and time-efficient solution of SR-related optimization problems. Accordingly, this paper proposes a vertex pruning approach that mixes local and backward search policies to achieve effective SR compression. In particular, the backward search strategy allows previously pruned vertices to be recycled when they are re-identified as critical. Subsequently, we incorporate this approach to enhance computational efficiency in the aforementioned applications. Numerical validations are conducted on two realistic medium-voltage Dutch grid cases. The results demonstrate at most 1.65% hypervolume loss with over 96.0\% SR data size reduction after applying the proposed compression approach. Empowered by this compression, the 8-dimensional (8D) SR becomes computationally tractable. Moreover, in relevant grid-aware computation challenges, the solving time can be reduced by at least 52.9% with at most an 1.36% optimality loss. All above validate both the efficacy and values of the proposed SR compression approach.

Editing multi-camera concerts requires coordinated decisions about musical relevance, visual quality and cut timing. We present a modular framework that combines instrument activity from multitrack audio, visible instruments and image sharpness with beat-based timing and probabilistic camera selection. An interpretable scoring model links these components to the choice of views over complete shot intervals. Component tests and an exploratory expert study on a four-camera concert yield mixed assessments across the editing variants. The findings motivate extending local audiovisual relevance with contextual and sequence-level criteria. The architecture provides a basis for testing these extensions and studying how component behaviour affects the quality of complete edits.

A robot policy needs to understand what a task is asking for in a scene, and how to move the robot to accomplish that task. To achieve these two skills, we introduce Zero-real Agent Policy (ZAP), a recipe that teaches a small off-the-shelf vision-language model to control a robot via a simple tool harness. We use a pretrained 27B parameter model to bootstrap world understanding and reasoning, followed by supervised fine-tuning and GRPO on simple simulated tasks to teach it how to move the robot. Without any real-world robot data during training, ZAP reaches 96.7% success on real-world tasks, matching GPT-6 Astra in our harness and outperforming the MolmoAct2 VLA (33.3%), while running 3x faster and 17x cheaper than GPT-6 Astra. We demonstrate that ZAP also generalizes to completely new objects, goals, strategies, and language by seating a test tube in a rack (and choosing a different hole in the rack when prompted), handing a test tube between arms mid-air, drawing an X on a tic-tac-toe board, and playing chess. These goals and types of manipulation don't appear in our training data. Furthermore, ZAP also transfers to an unseen embodiment. On a single-arm Franka, ZAP reaches 55.9% on RoboLab-120, 4.9 percentage points ahead of the leading VLA/WAM baseline.

Browser agents do not just need more practice. They need to practice the right things. We introduce MiniWeb, a realistic miniature of the web comprising 65 interconnected websites across 15 domains, where agents navigate, search, and transact as people do. Its 369 human-annotated tasks cover everyday workflows and label 37 reusable interaction primitives grouped into 7 skill families, allowing us to diagnose which skills fail rather than only whether a task succeeds. Benchmarking ten open-weight agents, we find that drag & gesture is consistently the weakest family, with no agent solving more than 24% of its instances. To address this gap, we introduce MiniWebGen, a framework that autonomously generates verified, skill-centric training data from MiniWeb. Fine-tuning Qwen3.5-4B on just 3.1K verified trajectories improves single-primitive success by 24.9pp across 13 unseen websites, with gains across every skill family. However, these trajectories teach short-horizon interactions, which do not directly translate to full-task success. We build on this result with MiniWebAgent, which pairs the trained executor with a planner that decomposes full tasks into the skill-level steps it learned to perform. This approach improves task success by 19.4pp points on unseen MiniWeb sites and transfers to external benchmarks, with gains of 9.3pp on WebArena-Lite, 16.1pp on Online-Mind2Web, and 20.0pp on WebVoyager, without training on their trajectories. We provide a simple recipe for compact browser agents: practice reusable skills directly, then compose them to solve unfamiliar tasks.

The small Solar System bodies (SSSBs) preserve records of the formation and evolution of the Solar System. In particular, the near-infrared wavelength region around 3 $μ$m provides crucial insights into the origin and transport of hydrated materials and organics, although it is difficult to observe from the ground because of strong telluric absorption. We assess the observability of SSSBs as complementary science to the ESA Ariel space mission, operating in the visible to mid-infrared wavelength range (0.5--7.8 $μ$m). Using the NASA JPL Horizons ephemeris system, we evaluate the visibility of known SSSBs from 2032 to 2035. Our analysis shows that Ariel can spectroscopically characterize up to 402 unique SSSBs under a conservative limiting visual magnitude of $V=13.6$. If observations can be extended to somewhat fainter targets ($V=16$), the number of observable SSSBs could increase to 2144. Assuming operational scenarios in which one SSSB is observed every 3, 7, or 30 days over a four-year mission, the total numbers of observable targets would be 357, 209, and 49, respectively. We further model the spectral energy distributions of candidate targets to evaluate their observational performance. We demonstrate that Ariel can resolve critical absorption features, such as those at 2.7 $μ$m and 3.4 $μ$m, with a higher spectral resolution than the previous infrared survey with AKARI. Observations of SSSBs with Ariel can provide insights into the distribution and evolution of life-forming materials in the Solar System and constitute a scientifically compelling component of Ariel's complementary science program.

Selection bias in large language models (LLMs) can persist despite instructions to ignore presentation order or prompts that elicit reasoning. Recent efforts to reproduce Jev-style System One behavior have produced a growing family of models that predict probabilities directly over supplied candidates. Yet systematic comparisons of selection bias across Jev versions, Jev reproductions, and generative LLMs remain limited in multi-option tasks. This leaves a central question: does Jev-style decision structure reduce selection bias? We address this question with a simple test that shuffles options while keeping question and candidate content fixed. Across two Jev versions, twelve Jev reproductions, and six generative LLMs, we trace individual answers through eight presentations and compare them with fixed-order repetition. On our primary collection of 1,700 questions from two benchmarks, Jev's accuracy varies by less than two percentage points, yet its cumulative failure set expands by 27--29\% relative to the original order, with first failures continuing into the final presentation. This loss of initial successes is substantially larger than under fixed-order repetition, even for highly repeatable Jev reproductions. Nor does Jev show a general advantage over the LLM baselines in preserving correct choices. Aggregation offers limited protection: voting yields modest gains for Jev, while even unanimous answers are frequently wrong. These results expose a gap between repeatable decisions and decisions that remain correct across equivalent presentations, showing why stable accuracy alone cannot establish resistance to selection bias.

Shapley-value-based explanations are widely used for feature attribution in machine learning. However, in computer vision, a feature may reduce the target-class score when added to the complete genuine image, even when the predicted label remains unchanged. We formalize such features as contextual distractors and investigate their attributions under semivalues and the least core. Under a weak average marginal monotonicity assumption, we derive quantitative semivalue--least-core gap bounds and sufficient conditions based on the probability of sampling smaller contexts. We further show that shifting semivalue weights toward larger contexts decreases the signed gap. Finally, controlled image-classification experiments and evaluations on Salient ImageNet provide strong empirical support for our theoretical findings.

A vulnerability costs more to fix the later it is discovered. As coding agents generate code faster than developers can review it, software needs a first line of defense screening each new function in real time. LLM agents have recently become capable of finding real-world vulnerabilities by working as a security expert would: they run analysis tools, explore the surrounding code, test hypotheses and weigh the evidence before forming their own judgement. This reasoning, however, spans many model calls, so that each verdict takes minutes of compute, far too long for real-time screening. Faster detectors, which fine-tune deep learning models or prompt an LLM for a direct verdict, are less accurate: they judge a function without the context an agent gathers, and an LLM's generated verdict is unreliable even when given some of it. We present REVA (REal-time Vulnerability Analysis), an AI vulnerability detector that brings the security context and code understanding components behind agentic detectors into a single forward pass of an LLM. For each new function, REVA gives the LLM a pre-computed threat model of its module, lightweight static analysis findings and call-graph information. REVA leverages a trained probe to read the LLM's internal representations of the new function in this context. On PrimeVul's held-out test set, REVA on GLM-5.3-Flash reaches 60.0% accuracy in about one second per function. This matches VulTriage (59.2%), the state-of-the-art agentic detector, run on the same LLM, while being over 250 times faster and less than a tenth as costly. On the same prompt, the probe is 14% to 17% more accurate than the LLM's own direct verdict. Across three open LLMs, REVA is at least as accurate as VulTriage, and about 10% more accurate on two of them. An LLM's internal representations thus encode much of what agentic detectors obtain through reasoning.

Frontier multimodal large language models (MLLMs) can control robots without robot-specific training, but how their capabilities compare and interact with post-trained vision-language-action (VLA) and world action model (WAM) policies across manipulation demands remains to be thoroughly evaluated. In this report, we evaluate GPT-6-Astra on EBench as the core of an agent that controls a robot through an agent harness under the same visual and proprioceptive observation protocol as post-trained policies. We study two agent configurations: direct control, in which the agent generates waypoint commands directly, and harness with policy, in which it can also invoke a post-trained VLA or WAM policy for action execution. Under direct control, the agent achieves competitive overall performance relative to post-trained policies. It leads on short-horizon tasks and exhibits the least performance variation across generalization dimensions, whereas it underperforms post-trained policies on long-horizon tasks, tabletop manipulation, and high-precision contact-rich tasks. In the harness with policy, invoking a post-trained VLA or WAM combines the complementary strengths of both components, raising overall task success to the level of the strongest compared post-trained policies while substantially reducing token cost. However, performance remains below the per-task upper bound of the standalone components due to suboptimal policy invocation, fixed-instruction execution without subgoals, and grasp disruption when the agent resumes control. Overall, our study characterizes the capability boundary of a frontier MLLM agent and suggests directions for improving future agent harness design. Project page: https://internrobotics.github.io/EBench-Astra-Tech-Report/

In federated learning (FL), secure aggregation enables a server to compute the sum of local model updates of $N$ clients, without revealing any information on the individual model updates to the server, beyond their sum. Different clients may be subject to heterogeneous bandwidth constraints, and consequently may choose different numbers of quantization levels to represent the information they send. Most secure aggregation works require the clients to project the values of their update parameters to a common finite group (or field), which does not directly accommodate heterogeneous precision constraints on their transmitted masked updates. Existing approaches of secure aggregation with heterogeneous quantization levels reveal partial aggregates of model-updates, and can reveal segments of individual model updates, under certain dropout and client-server collusion patterns. To prevent this privacy breach, we propose an information-theoretic secure aggregation scheme that tolerates client drop-outs, and up to $T$ semi-honest clients that collude with each other and with the server, under heterogeneous quantization at clients. Our protocol utilizes the $(T+1, N)$ threshold black-box secret sharing construction, where secret shares of the client updates, despite belonging to distinct algebraic groups owing to different quantization, are generated from identical linear integer systems.