Papers

Billion-parameter vision-language-action (VLA) policies run either onboard, consuming substantial power, or on remote servers, adding communication latency. To address these drawbacks and better balance latency and onboard energy consumption, we propose VLA-ULAP. It partitions inference across decision times, interleaving remote VLA calls with predictions from an Ultra-Lightweight Local Action Predictor (ULAP). A single ULAP has $\sim$7.4M parameters including the frozen vision encoder. It combines current views, proprioception, and executed action history to predict chunks in one pass. Trained independently, it requires no VLA hidden states, online verification, or server round trips. On Jetson Orin Nano, ULAP takes 20.7 ms and 0.122 J of idle-subtracted energy per inference, versus 289.3 ms and 40.46 J for GR00T on RTX A6000. Across four simulated base-policy/benchmark pairs, VLA-ULAP removes 45.3-77.3% of VLA calls while retaining 95.0-98.5% of baseline success rates at selected operating points. On VLA-JEPA, VLA-ULAP also surpasses local acceleration alternatives, using an estimated 49.4% less inference time and 51.5% less GPU energy per successful episode than ACT, and 77.5% less time and 80.7% less energy than SP-VLA at higher success rates. In physical SO-101 trials, it similarly removes 70.3-71.9% of VLA calls without observed success-rate loss at seen or held-out placements. Measured device costs imply 64.6-66.4% less inference time and 70.1-71.7% less idle-subtracted energy per successful episode at these call counts. Beyond these savings, faster responses help VLA-ULAP exceed $\pi_{0.5}$'s success rate by 11.0 and 15.5 percentage points (pp) on two tasks in latency-aware LIBERO-Safety simulation while approximately halving VLA calls.

Modern GPU kernels increasingly stress the instruction supply path, while fixed instruction containers can leave substantial footprint slack. This paper presents an automated encoding-synthesis framework that treats instruction layout as a constrained slot-assignment problem over a validated instruction-form field specification. The formulation separates semantic field identity from physical bit positions and supports tied, pinned, and free placement constraints, making it applicable when recurring decoded fields are not frozen by a public format contract. We instantiate the framework for NVIDIA SASS: raw public encoding text is normalized into a machine-readable specification, a SASS disassembler is validated against nvdisasm on 3.78M instructions, released as an open benchmark, and CP-SAT synthesis is used for fixed-length and variable-length encodings. On 142 Blackwell kernel inputs, variable-length synthesis reduces instruction footprint by 33%, with comparable reductions after re-synthesis on Ampere and Hopper; fixed-length synthesis on the same specification reduces decoder area by 16% against a decoder generated from the NVIDIA-observed 128-bit layout by the same generator and flow. Generated fetch/decode RTL meets 1.5 GHz in TSMC 22 nm with a replicated area delta of 0.12% of a GA100-class die; a same-node SRAM comparison shows the footprint reduction corresponds to about 9x this added logic in instruction-SRAM bit-cell area.

AI-powered development tools are now widely used to generate code and assist developers with routine programming tasks. Although existing work has identified vulnerabilities in AI-generated code, security-oriented work is often focused on vulnerability detection rather than risk assessment. To address this gap, this paper presents a Security Risk Assessment Framework (SRF) to evaluate the security risks of AI-generated code. SRF combines threat modeling, security analysis, and a quantitative risk evaluation approach based on vulnerability criticality. The framework is applied to a set of security-relevant programming tasks, where code generated by multiple AI-powered development tools is analyzed using Bandit and Semgrep. The results show that AI-generated code can introduce security vulnerabilities across all evaluated tools. They also show that risk levels vary by task type, as input processing and file handling tasks showed higher risk, while simpler tasks remained low-risk. Differences between tools exist but are smaller than differences across task categories. Overall, SRF enables reproducible evaluation of AI-generated code and provides a practical framework for assessing its security implications.

Common finite element discretizations for the incompressible Stokes equations with a continuous pressure approximation – the MINI element, the Taylor–Hood element, or stabilized equal-order elements – are not pressure-robust: the velocity error is polluted by the pressure best-approximation error, multiplied by the inverse viscosity. The established remedy replaces the test function in the momentum equation by a divergence-preserving reconstruction operator; for continuous pressure approximations this construction is cumbersome, requiring the solution of local problems (Lederer, Linke, Merdon & Schöberl, SIAM J. Numer. Anal., 2017). We propose a simpler alternative: interpolating the force by a Nédélec interpolation operator of suitable order instead of reconstructing the test function. Exploiting the commuting diagram property linking Nédélec and Lagrange interpolation operators, we show that the resulting method is pressure-robust for any $H^1$-conforming, continuous-pressure discretization. We give the complete analysis for the MINI element, and then for stabilized equal-order $P^kP^k$ elements of arbitrary degree $k$, proving that the interpolation-induced consistency error is pressure-robust and converges of optimal order. We validate the theory by several numerical examples.

Sign-based methods reduce communication costs in distributed environments, but aggregating local signs can introduce bias when data are heterogeneous. As a result, existing sign-based variance reduction methods fail to obtain the optimal convergence rates. In this paper, we solve this problem and obtain optimal rates for both nonconvex stochastic and finite-sum optimization. We first give a counterexample showing that majority voting can fail to approach stationary points even with exact local gradients. Motivated by this limitation, we propose tracking the global gradient at the server through unbiased compression of recursive gradient increments. As a result, we can obtain the convergence rates of $O(\sqrt{d/K}+\sqrt d (a/(nK))^{1/3})$ for the $\ell_1$-norm and $O(\sqrt{a/K}+\sqrt a/(nK)^{1/3})$ for the $\ell_2$-norm. Here, $K$ is the iteration number, $n$ is the number of workers, $d$ is the dimension, and $a=1+\omega$, with $\omega$ denoting the compressor's relative variance. For finite-sum problems with $M$ components, we combine periodic exact gradient refreshes with compressed component-gradient differences. The resulting total sample complexities are $O(M+d\sqrt{aM}\epsilon^{-2})$ and $O(M+a\sqrt M\ epsilon^{-2})$ for $\ell_1$ and $\ell_2$ gradient norms at most $\epsilon$, matching the corresponding bounds in centralized settings.

Quantum neural networks (QNNs) are typically built from variational quantum circuits (VQCs), which are limited by local measurements. Adaptive non-local observables (ANO) address this by jointly optimizing circuit parameters and multi-qubit measurements. However, existing ANO-based VQCs learn only a single static observable that remains invariant across all inputs. We propose QFWP-ANO, a novel architecture which employs a classical hypernetwork to dynamically program VQC parameters and/or non-local observables conditioned on each input. On multivariate time-series forecasting across four ETT datasets, QFWP-ANO achieves the lowest MSE in 16 of 20 settings and second-lowest in the remaining four, surpassing ANO-based and other strong baselines. On reinforcement learning tasks, QFWP-ANO consistently surpasses ANO-VQCs. Our results establish input-conditioned ANO as an effective approach for enhancing QNNs.

We develop and analyze a fully discrete, globally divergence-free hybridizable discontinuous Galerkin (HDG) method for a gradient-based Smagorinsky model. The method combines backward Euler time stepping, interior-penalty discretizations of molecular and nonlinear eddy diffusion, and an upwind convective flux. The discrete velocity is $H(\operatorname{div})$-conforming and pointwise divergence-free, which yields pressure robustness. For sufficiently large penalty parameters, we prove unconditional energy stability and existence of a discrete solution, and establish uniqueness under additional smallness conditions. A velocity error estimate is derived without explicit inverse powers of the molecular viscosity. The nonlinear facet residuals are controlled using local trace-approximation estimates and a viscosity-independent facet penalty. We retain the dependence of the discrete Gronwall factor on the filter scale and the mesh size; a mesh-uniform bound follows under suitable solution regularity, a fixed time-step margin, and the scaling $\delta=O(h)$ on quasi-uniform meshes. The reported manufactured-solution results are consistent with the resulting pre-asymptotic error bounds. Further flow examples illustrate the dissipative behavior of the method and are distinguished from the boundary conditions and parameter range covered by the analysis.

Generalist robot policies offer useful initialization, but refining compact specialists through limited physical interaction requires informative learning feedback. We present FIERCE, a generalist-initialized reinforcement learning framework centered on a unified, task-adaptive progress-failure evaluator. Its architecture shares an observation-language representation between an observed-progress head and an action-conditioned latent predictor whose past and current predictions feed a causal sequence head for task-failure estimation. Joint supervision from progress and preference labels, synchronized commands and observations, and terminal outcomes trains the evaluator; target-task rollouts support adaptation and calibration. Fixed evaluator snapshots provide progress shaping and failure-risk penalties alongside independently verified terminal rewards, while evaluator and policy updates alternate as new experience is collected. Refinement requires neither continued generalist action queries nor a dedicated target-task simulator or manually annotated dense rewards. Only the compact specialist is retained at deployment. The evaluation separates feedback quality, policy-learning efficiency, and deployment cost across simulation and two contact-rich real tasks. Code, model weights, and data-restoration tools are released at https://github.com/ar-mine/FIERCE.

Learning generalizable robot manipulation policies requires large-scale and diverse interaction data, yet collecting real-world demonstrations remains costly and difficult to scale. Existing approaches to data collection are either dependent on specific robot hardware that limits crowdsourcing and transferability, or suffer from incomplete annotation and limited behavioral diversity. Inspired by how games sustain long-term human engagement, we explore an alternative paradigm that turns data collection into an engaging gameplay experience and transfers the resulting human manipulation experience to real robots. We present Project Kitchen, a VR-based gamified egocentric data collection platform that elicits diverse, goal-directed manipulation while remaining independent of specific robot embodiments and hardware, making it applicable to broader and potentially large-scale deployment. To bridge the game-to-real gap, we further introduce Game2Policy, which extracts embodiment-invariant affordance cues, including contact points and sub-goal states, from gameplay trajectories. An affordance model is pre-trained on game-collected data and then jointly fine-tuned with downstream policies using only a handful of real-robot demonstrations. Experiments show that Game2Policy improves average success rates by 10.0 points in simulation and 18.3 points on real robots in the few-shot setting. User studies and quantitative analyses further show that Project Kitchen promotes diverse manipulation behaviors and provides an engaging data collection experience. These results demonstrate the potential of gamified virtual environments as a scalable source of manipulation knowledge. The platform and code will be released upon acceptance.

Warning: This paper contains examples of stereotypes and social bias. LLMs are increasingly used in interactive settings by the general public, making the evaluation of model behavior in multi-turn conversational scenarios important for safety, including stereotyping-related harms. However, existing multi-turn social bias evaluations often rely on pre-specified or template-based user inputs that do not adapt to model responses and typically assume a fixed dialogue length in advance. In this paper, we study social bias dynamics in response-conditioned multi-turn interactions using a controlled evaluation protocol that generates follow-up user queries from the evolving dialogue history and allows evaluation over variable numbers of turns. Experimental results show that LLMs exhibit social bias even in coherent, response-conditioned multi-turn interactions, revealing late-emerging bias, non-monotonic bias patterns, and bias re-emergence. These results motivate evaluations that extend beyond fixed-turn, pre-scripted protocols. Our findings highlight the importance of analyzing social bias as a turn-level dynamic phenomenon.

Fallacy-detection benchmarks pair fallacy classes with a single "valid" or "none" class that takes everything data collection did not label as a fallacy. A detector has two jobs, deciding whether an argument is fallacious and naming which fallacy it commits, and the false-positive rate is meant to measure the first. We show that what these benchmarks actually score is scheme recognition, the ability behind the second job. Their own test sets already show it: when a classifier misses a fallacy, the error lands on "none" rather than on another fallacy type, so detection is failing while classification holds. The reason is what the valid class lacks. The negatives that separate the two jobs are correct arguments using the same argumentation scheme as a fallacy, and they are scarce: nearly absent from the four benchmarks we examined, and rare even under deliberate search. A detector is therefore never tested where recognizing a scheme and judging its use come apart, and can pass on recognition alone. We construct the missing arguments, together with a control condition from the same pipeline that differs only in scheme, so whatever generation contributes, it contributes to both. The classifier labels the scheme-matched negatives as the source fallacy, and labels the wrong-scheme negatives as the scheme they actually use 85.9% of the time and as the source type 0.4%. The classifier has learned which scheme an argument uses, not whether it uses it correctly. The over-flagging follows: a model that scores 16.6% on CoCoLoFa's own valid class flags 58.9% of the constructed arguments. The same dissociation appears in three zero-shot LLM detectors that never saw these benchmarks. We release the items as Scheme Foils. A reported false-positive rate should not be trusted as a measure of detection until the valid class has been audited for scheme-matched coverage.

Large language models (LLMs) often suffer from capability stagnation in self-improvement training because fixed difficulty levels fail to adapt to their evolving proficiency. To address this issue, we propose STRETCH (Self-Taught Reasoning Evolution via Targeted CHallenge), a unified framework inspired by cognitive scaffolding theory. STRETCH introduces a dynamic Stretch Zone mechanism that continuously aligns question difficulty with the model's solving capability. Within a single parameter space, the model alternates between a Scaffolder that generates adaptive, boundary-pushing challenges and a Learner that that optimizes its solving trajectories through reinforcement learning. This dual-loop co-evolution effectively stabilizes training, mitigates reward hacking and promote progressive reasoning growth. Experiments on both negotiation and operation research benchmarks demonstrate that STRETCH consistently outperforms strong prompting and domain-specific baselines. Further scaffolder configuration analysis shows that dynamic difficulty alignment is critical for sustained capability improvement and synchronized reasoning evolution.

Emergency management assistance programs, such as relief distribution, are essential for delivering necessary supplies to affected communities. However, these programs operate in a decentralized network of local centers that face uncertain local demand and supply dynamics, resulting in inconsistent avail- ability of local services. Redistribution of supplies among these local centers reduces these imbalances, but the centers often make decisions independently, with limited information and disrupted transportation. This study develops CoRe-MARL, a cooperative multi-agent reinforcement learning (MARL) framework, by formulating a decentralized partially observable Markov decision process (Dec-POMDP). We treat each center as an agent that learns a redistribution policy to improve the service in the worst-case region and reduce the service gap across regions while protecting network-wide service. We incorporate a recurrent network that captures evolving supply and demand dynamics without direct observation, while multi-agent proximal policy optimization (MAPPO) enables centralized training and decentralized execution (CTDE). We evaluate the framework in a simulated environment with diverse trajectories, where exact dynamics are not observed by actors and the MAPPO critic. We compare the recurrent MAPPO with the recurrent independent PPO (IPPO) and a local only heuristic, and find that MAPPO reduces the service gap across local centers and enhances service for the worst-served center while maintaining competitive network-wide service. The recurrent MAPPO also shows consistent performance across diverse trajectory patterns, demonstrating its ability to adapt to evolving dynamics. The findings demonstrate the capability of cooperative learning for decentralized redistribution and improving equitable service under uncertain and evolving dynamics.

The main contribution of this work is to identify model parameters for an Anion Exchange Membrane (AEM) system directly from measured operational data, enabling their use in power system studies. Due to the relatively low technology readiness level (TRL) of AEM electrolysers, only limited literature reports operational parameters supported by openly available experimental data. By capturing the realistic dynamic behaviour and power consumption of the AEM electrolyser, this model allows for more accurate evaluation of system topologies and control strategies for green hydrogen production. This paper presents a model of an AEM electrolyser based on experimental data from Enapter modules installed in the Power-to-X (PtX) Laboratory at the Technical University of Denmark [1]. The model comprises a polarisation curve and a model of current dynamics during operation. It is validated against operational data and can be applied in grid-balancing studies as well as in assessments of how electrolyser dynamics influence the power system. The results suggest that the AEM technology is suitable for services such as smoothing wind farm power output and provide system services to support frequency balancing. The findings highlight the importance of considering the constraints of the process when designing for flexible load operation, and consideration of the interaction between the power control and process control systems.

The reuse of metal powders in directed energy deposition using laser powder offers promising sustainability benefits for additive manufacturing, yet its impact on part quality remains unknown. This study investigates the influence of powder reuse on the directed energy deposition process of Inconel 625. Virgin powder was first characterized in terms of flowability, morphology, size and chemical composition, then used to manufacture parts under fixed process parameters. These parts were analysed through two configurations (as-built and hot isostatically pressed) to assess microstructural evolution and mechanical performance. Results show that hot isostatic pressing reduces porosity and promotes grain recrystallization, enhancing ductility while decreasing tensile strength. Recovered powders from one manufacturing cycle, were also characterized. Particle size distribution and morphology remained stable, interstitial contents, particularly oxygen, increased, potentially affecting laser-material interaction and melt pool behaviour in future production. These findings highlight the importance of evaluating powder reuse through a full-process approach, from powder to final part. Future work will focus on the tensile and fatigue performance of parts produced with recovered powders.

We study the canonical Maximum Clique and Maximum Independent Set problems in the one-pass edge-arrival graph streaming setting. Here, the edges of some input graph $G = (V,E)$ are presented one at a time (possibly including deletions), before an algorithm needs to produce either a large clique or independent set at the end of the stream, with the focus being on space complexity. We are interested in finding $\beta$-approximate solutions, for any $\beta \geq 1$. Previous work gave an algorithm using $\tilde{O}\left(n^2/\beta^2\right)$ bits of space, together with a corresponding $\tilde{\Omega}\left(n^2/\beta^2\right)$ two-party communication lower bound [Halldórsson et al., ICALP'12], seeming to resolve the problem. However, their algorithm crucially relies on randomness, and the best known deterministic algorithm remains a folklore derandomisation using $O\left(n^2/\beta\right)$ bits of space, leaving a (deterministic) gap of size $\tilde{O}(\beta)$. We make progress towards this gap with a strong lower bound for a restricted range of $\beta$: whenever $\beta \leq \sqrt{n}/36\log n$, any deterministic algorithm for either problem must use $\Omega\left(\frac{n^2}{\beta\cdot\log n}\right)$ bits of space. Our proof is via a two-party one-way communication lower bound, and highlights the power of randomness when approaching either of these problems.

Video streaming dominates global internet traffic, yet conventional pipelines remain inefficient for structured, human-centric content such as sports, performance, or interactive media. Standard codecs re-encode entire frames, foreground and background alike, treating all pixels uniformly and ignoring the semantic structure of the scene. This leads to significant bandwidth waste, particularly in scenarios where backgrounds are static and motion is constrained to a few salient actors. We introduce GenStream, a semantic streaming framework that replaces dense video frames with compact, structured metadata. Instead of transmitting pixels, GenStream encodes each scene as a combination of skeletal keypoints, camera viewpoint parameters, and a static 3D background model. These elements are transmitted to the client, where a generative model reconstructs photorealistic human figures and composites them into the 3D scene from the original viewpoint. This paradigm enables extreme compression, achieving over 99.9% bandwidth reduction compared to HEVC for the continuous data stream. We partially validate GenStream on Olympic figure skating footage and demonstrate potential for high perceptual fidelity under minimal data. While acknowledging the significant computational costs shifted to the client and challenges in generalization, GenStream opens new directions in volumetric avatar synthesis, canonical 3D actor fusion across views, and personalized viewing experiences, laying the groundwork for scalable, intelligent streaming in the post-codec era.

Cloud-native microservices architectures rely on network namespaces for isolation, with the overhead of container communications remaining a critical performance bottleneck. While colocating containers on the same host mitigates some of this overhead, it cannot match the performance of communication within a single network namespace. Existing solutions either require application rewrites or fail to support the full Linux network stack expected by containerized applications. In this paper, we present netkit, an eBPF-based datapath that specializes the Linux networking stack to eliminate redundant backlog queue traversals during network namespace transitions. netkit leverages eBPF to transparently redirect packets between namespaces, bypassing unnecessary buffering while preserving compatibility with existing container applications. Our implementation in the Linux kernel, integrated with minimal changes to the Cilium network plugin for Kubernetes, improves throughput by up to 37% and achieves parity between container-to-container and process-to-process communications, effectively closing the performance gap introduced by namespace isolation.

Multi-modal talking avatar synthesis aims to generate realistic talking videos from a reference portrait and speech. Despite rapid progress in diffusion-based methods, existing approaches still struggle to jointly achieve accurate lip articulation, human-preferred motion aesthetics, and efficient inference. We observe that phonetic accuracy and motion aesthetics arise from fundamentally different sources and should be addressed at complementary stages rather than learned implicitly by a single generator. Based on this insight, we propose VibeAvatar, which disentangles these two objectives through a Phonetic Kinematics Adapter (PKA) that converts recognition-oriented speech features into phonetic-kinematic conditions at the conditioning stage, and an Aesthetic Motion Policy (AMP) that optimizes a flow-consistent stochastic sampling policy via Group Relative Policy Optimization (GRPO) at the post-training stage. With a lightweight flow-based motion generator operating in a compact 1D warp-based latent motion space, VibeAvatar achieves state-of-the-art results in articulation, aesthetics, and efficiency on both objective metrics and user studies, while generating a 10-second 512px video in under 10 seconds with only $\sim$3GB VRAM.

Visual-inertial odometry (VIO) is a core capability for autonomous operation in GPS-denied subterranean environments, yet its reliability can degrade sharply under sensor drift, calibration errors, and dynamic occlusion. Existing evaluations mainly emphasize nominal-condition accuracy, offering limited insight into when practical deployment failures occur. In this work, we present a failure-centric stress-test benchmark for VIO in underground environments using the CERBERUS dataset. We systematically evaluate four representative VIO systems spanning filtering-, optimization-, and learning-based paradigms under nine practical perturbation settings, including IMU bias and noise variation, camera intrinsic and extrinsic drift, and dynamic scene occlusion. Beyond conventional trajectory error, we analyze robustness limits through coverage ratio and failure thresholds, revealing breakdown behaviors that are not captured by nominal-condition performance alone. Our study shows distinct vulnerability patterns across VIO paradigms: some methods are more sensitive to inertial degradation, while others are more affected by geometric miscalibration or dynamic interference. These results provide deployment-oriented guidance for VIO selection, calibration prioritization, and reliable operation in challenging underground scenarios. To support reproducible evaluation and future extensions, we will release the full benchmark scripts and evaluation pipeline.

3D Gaussian Splatting (3DGS) enables photorealistic novel view synthesis, but transmitting gigabyte-scale scene data remains challenging for immersive applications. Traditional HTTP Adaptive Streaming over TCP introduces Head-of-Line (HOL) blocking and coarse segmenting ill-suited to fine-grained 3DGS delivery. We propose MoQSplat, which maps 3DGS content onto the Media over QUIC (MoQ) transport hierarchy. MoQSplat partitions scenes into spatial Tracks, clusters splats into semantically coherent Groups, and constructs progressive-quality Subgroups mapped to independent QUIC streams to eliminate connection-level HOL blocking. Using a stateless, subscriber-driven adaptation loop, clients dynamically request spatial regions and quality tiers based on six degrees of freedom (6-DoF) frustum visibility, distance, and foveal alignment. We evaluate the core components on a prototype implementation, showing that opacity-based pruning outperforms scale-based pruning for progressive delivery. The source code is available at https://github.com/emanuele-artioli/MoQSplat.

State-of-the-art vision-language-action models (VLA) for autonomous driving face critical limitations: excessive parameter counts, inefficient high-resolution image processing, and lack of temporal memory. We introduce Fast and EffectIVE VLA (FIVE-VLA) to address these through two key contributions. First, we employ an efficient vision encoder that processes high-resolution ($448 \times 896$) images while generating only 98 tokens, over $5\times$ fewer than existing approaches, and bypass text generation entirely for single-pass trajectory prediction. Second, we propose Recurrent Action Memory (RAM), a lightweight module that conditions action prediction on previous action tokens, providing temporal context critical for manoeuvres such as overtaking and emergency braking. With only 641M parameters, FIVE-VLA completes $\sim$10% more routes without traffic rule infractions than the previous state-of-the-art VLA on the challenging Bench2Drive closed-loop driving benchmark. Non-reactive open-loop simulation on the large-scale real-world NVIDIA Physical AI AV dataset shows 10.2% and 7.7% lower collision-violation rates than SimLingo in single- and four-view settings, respectively. Additionally, FIVE-VLA runs at $\sim$30 fps on an A100 and $\sim$4 fps on a T4 GPU (proxy to an edge device), representing an 8-30$\times$ speedup over previous methods.

Classifiers can make identical predictions yet require labels to compare their selective performance: confidence ranks weight the same errors differently. We quantify this requirement for the area under the generalized risk-coverage curve (AUGRC). A prelabel lower bound rules out insufficient budgets. With all labels known, a covering linear program bounds the minimum number of labels sufficient to fix the winner (the certificate size) within $K-1$ labels for $K$ candidates. For fixed $K$, independent uniform orders and identical predictions, the prelabel bound approaches one quarter of the pool. With iid Bernoulli errors independent of the orders, every exact acquisition policy reads almost all labels asymptotically, although a two-candidate certificate needs only half. Across 108 feature-panel comparisons on nine datasets, disagreement labels settle every accuracy choice but no AUGRC choice. A 20% budget is ruled out in 96 conditions; certificates need 56-57% on average. On ten conditions with pretrained image classifiers, confidence-score choice reads 68-91% of 10,000 labels for exact selection and 50-67% with AUGRC tolerance $5\times10^{-4}$. An exact stopping test works with any acquisition order. Together, these results link confidence ranks to label budgets and certified model comparison.

Creating deformable assets for robot manipulation requires jointly specifying their geometry, appearance, and physical properties. This is especially challenging for deformable objects, since text and images provide limited evidence about how they deform and respond to contact, yet these responses directly affect their suitability for interaction. Automated generation therefore needs to resolve coupled physical requirements and use interaction evidence to guide construction and refinement. We present DeformSmith, a framework that enables automated generation of interactive, physically credible deformable assets from text or a single image. Through hierarchical agentic construction and a shared physics-grounded harness, it progressively builds, tests, and refines geometry, physical models, material behavior, and robot interaction until the resulting asset is ready for simulation and manipulation. Robot interaction closes the generation loop through manipulation feedback and replayable interaction data. Results show that DeformSmith generates assets with better visual quality and physical plausibility than state-of-the-art baselines, including PhysGen3D, PhysGM, and PhysX-Omni, while supporting the synthesis of data for robotic manipulation of deformable objects. Project page: https://can-lee.github.io/deformsmith-web/

Subspace methods such as multiple signal classification (MUSIC) achieve super-resolution direction of arrival (DoA) estimation by exploiting the orthogonality between the array manifold and the noise subspace of the measurements. Their accuracy therefore depends on the assumed manifold and degrades under model mismatch, while parameters not identifiable from the spatial manifold cannot be recovered. In this work, we propose the conditional neural manifold (CNM), which replaces the fixed manifold with an observation-conditioned mapping from source parameters to steering vectors. An encoder maps the snapshots to a latent scene representation that conditions a zero-initialized neural field over the parameter space. The manifold is learned without steering-vector supervision by shaping the resulting MUSIC landscape. Since the correction acts on the manifold rather than on the estimator, it can be used by other manifold-based methods without modification. The CNM restores resolution under array imperfections, colored noise, correlated sources, and near-field propagation, and resolves the angle-frequency ambiguity inherent to the nominal spatial manifold.

We analyze the learned input-output behavior of GLU-based neurons in large language models (LLMs). We propose a simple analysis method: For each neuron, we compute the cosine similarities between its input (reading) and output (writing) weight vectors. In this scheme, a strong negative cosine similarity indicates the neuron weakens the direction it detects in the residual stream, so we call this a weakening neuron. This allows us to gain a number of novel insights. First, we show that nine different LLMs have similar patterns: weakening neurons appear mostly in late layers whereas their counterparts, (conditional) strengthening neurons, are frequent in early-middle layers. Second, we find that weakening neurons display surprising behavior: even though there are few, they activate often and have a large influence on model behavior. Third, weakening neurons have a strong effect on model output when gate values are negative – which is surprising since negative gate values are not expected to encode functionality.

Classical dynamic programming represents optimal sequential decisions through value functions and policies. While this functional representation is natural for computing optimal decisions, it does not directly identify the mathematical object governing policy reconstruction, representation complexity, or oracle-query complexity once an optimal policy is fixed. This paper addresses this question by developing a geometric theory of structured optimal policies in which the decision-boundary geometry induced by the policy becomes the primary object of analysis. We show that, under suitable structural regularity conditions, this geometry provides the minimal representation required for policy reconstruction and determines the statistical and computational complexity of the reconstruction problem. Building upon this representation, we establish structural properties of policy-induced decision geometry, introduce intrinsic notions of boundary and decision complexity, derive information-theoretic measures of decision compression, and obtain statistical guarantees for boundary estimation and policy reconstruction from black-box policy queries. Collectively, these results demonstrate that, for the structured decision problems considered here, the complexity of policy reconstruction is governed by the geometry of the decision boundary rather than by the cardinality of the ambient state space. Controlled numerical experiments examine the principal theoretical predictions and provide empirical evidence consistent with the proposed framework.

As corporate AI adoption continues to grow, enterprise-grade LLM agents are being deployed into sensitive contexts such as hiring, healthcare, and finance. In these contexts, compliance with rules specified in an agent's system context is a first-order legal concern. Currently, no evaluation framework systematically measures which LLM models tend to violate compliance rules, especially under pressure from a persistent user, a hurried manager, or circumstances where violation is convenient or attractive. We introduce PACT (Pressure-Applied Compliance Testing), a benchmark for rule-following under pressure in AI agents assisting employees in daily tasks across twelve regulated enterprise domains and forty-eight scenarios, each set in a realistic multi-turn conversation. Each benchmark item pairs a standing rule against a rule-violating shortcut, and applies a battery of pressures across different wordings and system-prompt modes. We construct PACT component by component under strict LLM-as-judge auditing to ensure samples are unambiguous, ungameable, and realistic enough to avoid eliciting evaluation-aware behavior. We use PACT to profile LLM compliance across six complementary metrics that create a holistic picture of an AI assistant's robustness under pressure and throughout multi-turn conversations, its transparency, and ability to correctly discern where a rule applies. We aggregate this profile into PACTScore, a reliability-weighted compliance rate over all items and modes. Our results across 22 common LLM models spanning multiple providers and sizes show substantial variability in compliance across models and metric dimensions. Even the strongest assistants mis-apply a rule on 6 to 10% of items, and ordinary user pressure raises the violation rate by 65% on average. PACT highlights compliance risks in LLM assistants, motivating guardrails and careful model selection.

Despite recent progress in learned image compression, existing methods remain computationally expensive on resource-constrained hardware, particularly CPUs. We introduce PULSE, a practical codec that enables (1) low-latency decoding on diverse hardware platforms with an ultra-low-complexity 5.2 kMAC/pixel neural receiver, and (2) efficient bit-exact entropy coding with an integer linear CDF predictor and a meta prior. To recover compression performance under this tight budget, we introduce an agentic evolution process guided by heuristic probes that iteratively improves the architecture through human-LLM collaboration. PULSE decodes a 1080p image in 126 ms on a single CPU thread while achieving compression performance comparable to HM. After perceptual optimization, PULSE competes with larger perceptual codecs like MS-ILLM. Codes are at https://github.com/microsoft/GenCodec/tree/main/PULSE

A piezoelectric transducer embedded in a pitch-plunge aerofoil with a finite-mass trailingedge flap converts the limit-cycle oscillation that follows flutter into electrical power. The design findings for such a harvester have so far been obtained with incompressible strip theory, whose aerodynamic state is a handful of lag variables. Here the same section, the same transducer and the same test cases are coupled instead to the two-dimensional compressible Euler equations, giving a full-order model of twelve thousand states, and the two aerodynamic models are placed side by side on the flutter boundary, on the limit cycle and on the harvested power, with the strip-theory results compared against higher fidelity aerodynamic modelling provided by CFD. The degree of freedom carrying the transducer remains the first-order design variable, and the ranking of the mountings and the position of the optimum in the coupling and load-resistance plane are unchanged, while strip theory is found to underpredict the harvested power of the compressible section and to overstate the effect of the transducer on stability where the electrical time constant meets the flutter frequency. A nonlinear reduced model of four states, built on eigenvectors of the coupled Jacobian at the velocity of interest, reproduces the frequency of the cycle, its pitch amplitude, the transducer voltage and the mean power, and misstates the proportion of plunge to pitch in the orbit and its phase. That error is shown to be independent of the amplitude of the cycle and to survive enlargement of the basis, which places it in the retained subspace rather than in the order of the expansion, and suggests that such models should be judged on the shape of the orbit rather than on a single amplitude.