Papers

World models are useful for robotic manipulation because robots can predict how actions change the states of objects before executing them. We present PointCast, a point-set world model that spans rigid, articulated, and deformable object manipulation. Its state is a set of 3D points on the object and the end-effector, mesh-free and topology-agnostic. Each point keeps its identity and is supervised on its own trajectory, which teaches the model where every point goes rather than only the shape the points form. Its backbone is a diffusion transformer that denoises a short window of future point positions, conditioned on the points' recent history and the commanded end-effector motion. The backbone's attention alternates between local and global, and cross-attention to the end-effector carries the coupling. This one architecture at 19.8M parameters and one training recipe cover four regimes, rigid objects, cloth, rope, and multi-joint cabinets, with a separate checkpoint trained for each. Trained on randomized simulation and scored against four baselines on the same metric, it is best on three of four regimes and second on rigid. Trained on a real-world robot teleoperation dataset, it has the lowest mean error in four of its six categories, is second in the other two, and improves on the dataset's own model in all six; zero-shot, its simulation checkpoints are best on two of four captures. Frozen inside sampling-based model-predictive control at one network evaluation per window, it plans four simulated tasks over 64 episodes, competitive with or outperforming every baseline on each. Project website at https://pointcast-wm.github.io.

Score matching has driven major advances in classical generative learning by enabling models to learn from data without evaluating intractable normalization constants, or partition functions. Yet, extending this principle to quantum learning requires rethinking its foundations, as quantum states are described by noncommuting density operators rather than scalar probabilities. The noncommutativity creates fundamental challenges not only in defining quantum scores, but also in developing a training framework with efficient circuit implementations and rigorous theoretical guarantees. In this work, we bridge this gap by establishing a general quantum score-matching framework with end-to-end theoretical guarantees. Applied to Gibbs-state learning, our approach avoids additional thermal-state preparation and achieves information-theoretically optimal sample complexity in the high-temperature regime for Hamiltonians with bounded locality and interaction degree. This positions score matching as a new route to state-of-the-art performance in learning quantum Gibbs states. Beyond these theoretical results, numerical simulations show that our method remains effective even when gradients are estimated inaccurately under limited measurement budgets. Experiments on IBM quantum hardware further demonstrate that quantum score matching is NISQ-friendly: without any error mitigation or correction, it reduces the relative Hamiltonian-parameter error from 64% to approximately 10%. Together, these results extend score matching into an experimentally realizable paradigm for quantum-state learning.

Social simulation studies how individual behavior and social interaction produce collective outcomes. In social media marketing, campaign actions shape which consumers encounter the content and how they respond; these responses then spread through the population. We propose OranSim, a social simulation framework that connects creative, creator, targeting, and budget choices to this process. Heterogeneous consumers receive exposure according to content matching and platform allocation and generate initial responses, which propagate among 60 population segments. Candidate campaigns share the initial population and aligned random numbers, making their response trajectories comparable under action changes. In a controlled synthetic campaign, doubling the budget approximately doubles reach while lowering mean content match and engagement probability among the reached consumers; mean 14-day cumulative simulated response mass rises to 1.96 times the baseline. LightGBM predictors fitted to 39,000 historical RedNote notes estimate platform engagement with log-scale $R^2$ of 0.56–0.62 in five-fold cross-validation; a separate 12,154-note corpus supplies temporal, unseen-creator, and held-out-niche test splits. Public-data experiments evaluate policy-value estimation and audience ranking, and paired synthetic outcomes test counterfactual scoring. Together, scenario trajectories and engagement estimates support campaign selection according to a prespecified marketing objective. Code is available at https://github.com/OranAi-Ltd/oransim.

Reinforcement learning with verifiable rewards (RLVR) supervises mathematical reasoning through final-answer correctness, but provides little guidance on individual tokens. On-policy distillation (OPD) supplies dense feedback on student-generated responses, yet teacher preference need not reflect correctness. Recent hybrids combine OPD and verifier-derived advantages or reweight task credit using teacher ratios. However, teacher guidance enters after verifier-based group normalization, and token reweighting need not preserve the total task credit assigned to each response. We introduce Unified Entropy-Calibrated Credit Redistribution for GRPO (UECR-GRPO), which integrates verifier and teacher signals within a single GRPO-style update at both the response and token levels. Path-Utility Unification (PUU) combines verifier reward and a teacher-to-anchor path log-ratio in a single KL-regularized objective. Its on-policy implementation uses a length-normalized teacher score and combines both rewards before group normalization and PPO clipping, allowing teacher evidence to influence the response ranking. Entropy-Calibrated Redistribution (ECR) then uses the signed teacher–old-policy token gap to redistribute the verifier-derived component. Full-vocabulary teacher entropy attenuates uncertain guidance, while a response-wise zero-sum projection preserves the total task credit and its token-wise sign before clipping. Across five mathematical reasoning benchmarks, UECR-GRPO achieves average \(\mathrm{Avg@12}\) accuracies of 17.21% and 65.09% with Qwen3-1.7B and Qwen3-4B students, respectively, exceeding the strongest baseline at each scale by 0.89 and 0.56 percentage points.

Watch time is the primary engagement signal in short video feeds, and its prediction directly affects ranking and exposure. Existing methods improve watch time prediction by correcting duration bias or modeling richer distributions, but most expose only an expected or debiased watch time at serving time. Even when video duration is available to later models, the interface gives only one estimate of watch time and no probabilities for completion, overplay, or other regions relevant to downstream tasks. To address this limitation, we propose the Distributional Serving Interface (DSI), which has a distribution provider, a compact, low-dimensional summary, and lightweight readouts tailored to each task. The provider learns a joint distribution over four watch states derived from watch ratio and their event times; rules based on video duration remove incompatible combinations, while a restoration loss preserves accuracy in seconds. The summary reduces this distribution to a small set of event probabilities, time scales relative to duration, and uncertainty statistics. After training the provider, we fix its parameters and train value and ranking readouts that combine the summary with raw context. Across KuaiRec, KuaiRand-1K, and WeChat21, the complete DSI system achieves the lowest MAE on all three datasets, beating the strongest result among nine baselines by 1.9% to 8.5%, and achieves the best XAUC on two. It also leads retrieval metrics that account for video duration when complete systems are compared. With matched readouts held constant, the summary retains information relevant to each task beyond a predicted mean paired with video duration. Using the same lightweight linear heads for each new target, it also performs best on two new watch-time targets and improves a separately logged engagement target, while a randomly initialized provider does not reproduce this gain.

Terminal intercept-angle control against a maneuvering target can substantially increase the required missile acceleration, potentially leading to saturation and interception failure unless acceleration limits are explicitly addressed. The engagement is therefore formulated as a linear-quadratic optimal-control problem with bounded acceleration commands. Polynomial approximations of the line-of-sight projection coefficients are used to better represent the nonlinear engagement geometry and estimate the time-to-go. The bounded optimal command is derived over saturated and unsaturated arcs, whose switching times are computed at each guidance step. The guidance law is derived for arbitrary linear missile dynamics and implemented for zero-order missile dynamics. For the zero-order model, the conditions under which the terminal demands can be met are derived in closed form, yielding the minimum and maximum reachable commanded terminal intercept angles. Performance is evaluated in nonlinear simulations. Compared with its unconstrained counterparts, the bounded formulation yields substantially smaller miss distances and terminal-angle errors when saturation is encountered. Unlike corresponding bounded miss-only guidance laws, the proposed law does not reduce to its unconstrained counterpart for minimum-phase missile dynamics because the acceleration command can saturate near the end of challenging engagements. The bounded law anticipates this saturation and compensates through earlier maneuvers.

Humanoid control, leveraging human demonstrations, has achieved diverse, agile, and natural locomotion behaviors through reinforcement learning (RL). While this paradigm has yielded remarkable performance in physical humanoid control, how to eliminate specific motions from learned policies remains insufficiently explored. Addressing this issue is motivated by pressing safety and privacy concerns: the removal of malicious, poisoned, or suboptimal motions, as well as copyright-protected motions subject to the right to be forgotten under regulations such as the GDPR, is of critical importance. To this end, we propose {ForgetMimic}, the first motion-level unlearning method designed specifically for physical-world humanoid control. The core idea of ForgetMimic is as follows: given a policy $\pi_\theta$ trained on $N$ motions, our method degrades performance on a target subset of $K$ motions while preserving the effectiveness of the remaining $N-K$ motions. Furthermore, we identify and resolve two key training mechanisms in robot control that lead to unlearning failure. We conduct extensive experiments on the Unitree G1 and H2 humanoid robots across 12 motions, including Dance, Fight, Flip, and others. Experimental results demonstrate that ForgetMimic effectively eliminates memory of designated motions while maintaining the normal operation of all other motions.

Optimization problems (OPs) are key to solving many challenging research problems in robotics. However, reproducibility still remains a major issue. In this paper, we present Amplify, a lightweight nonlinear programming library aimed at reproducible results of robotic-related trajectory optimization problems. The minimalistic requirements for the 537-line library (80 characters per line) are an Internet connection, familiarity with the AMPL modeling language, and a text editor. Our primary contribution is the formulation of a library where trajectory optimization algorithms are represented directly within the optimization model. Specifically, we implement the algorithms used to compute the dynamics, trajectories, and reference motions as constraints of the OP in a declarative programming paradigm. We outline how our formulation of objectives, decisions variables, and constraints can be implemented in other transcription libraries that want to be lightweight and reproducible. We also compare the Amplify framework with 3 other libraries across examples of benchmark optimization problems across several fields, including bipedal locomotion and grasp planning.

Consumers increasingly delegate purchasing decisions to Large Language Models (LLMs) acting as surrogate consumers. Using "Tool-Lab," an adaptation of information-board process tracing that places product attributes behind costly tool calls, we examine how marketing pricing cues (i.e., just-below pricing and promotional framing) influence AI shopping agents. Across eight commercially deployed LLMs from three providers, we trace pre-choice information acquisition. Under zero cost, pricing cues rarely mislead. Imposing acquisition costs under a vague goal prompt leads LLMs to omit diagnostic attributes required to compute unit price and choose suboptimal choices resembling human heuristics. Relative to a specific goal prompt that mainly preserves diagnostic search and choice optimality, a vague goal prompt under constraints creates a search-mediated vulnerability. This research demonstrates that marketing heuristics in delegated AI shopping are governed by storefront information architecture, not necessarily immutable LLM flaws.

Vision-language models (VLMs) offer a promising approach to long-tail autonomous driving, but existing driving datasets provide limited supervision for connecting decision-critical visual evidence with reasoning and planning. We introduce AnchorReasoning, a visually grounded reasoning dataset built on WOD-E2E, containing 416,119 annotated frames and 395,379 decision-critical elements across four major categories and 19 fine-grained types. Each frame is organized as a visually grounded chain-of-thought (VG-CoT) that links decision-critical element identification and localization, element attributes and implications, driving-action rationale, and action and trajectory planning. We further develop a curriculum supervised fine-tuning strategy that progressively learns these hierarchical capabilities, together with an object-size-aware grounding metric for evaluating localization quality. Experiments across eight general-purpose, embodied-AI, and AV-specific backbones show that VG-CoT supervision improves grounded reasoning and trajectory prediction. Across models, 5-s ADE and FDE decrease by 7.84 and 11.86, while RFS Frame and Cluster improve by 1.66 and 1.70. These gains are achieved with 18.5 fewer reasoning tokens and 0.32 s/frame lower inference latency on average, demonstrating the value of visually grounded, decision-focused supervision for VLM reasoning and planning in long-tail autonomous driving.

Control barrier functions (CBF) are a popular safety filter to ensure safety for nonlinear dynamical systems. However, when the system is subject to uncertainties and disturbances, this requires the use of robust variants of CBFs, which can be difficult to construct and can be overly conservative, especially for high-dimensional systems under input constraints. In this work, we propose a new approach to solve these challenges by introducing Least-Effort Adversarial Potentials (LEAP), a certificate that quantifies the robustness of a given state against disturbances in terms of the effort required by the disturbance to cause failure. We show that LEAP is a CBF for the undisturbed system, but can also be used to construct a safety filter that is robust to disturbances whose cumulative effort is bounded. We propose a method for constructing LEAPs with on-policy deep reinforcement learning. Next, we demonstrate LEAPs in simulation on a variety of multi-agent systems with disturbances and uncertainties. Finally, hardware experiments on a quadruped and quadrotors validate that LEAPs are well suited to tackle the disturbances and uncertainties from real-world robotic systems.

We develop a curvature-domain formulation for continuous-aperture signaling, in which the transmit phase is represented through its second spatial derivative after quotienting out affine piston-and-tilt gauge freedom. The resulting gauge-fixed synthesis operator is bounded and compact, with sharp Poincare-Wirtinger constant $C_L=L^2/\beta_1^2$, and its modal Gram spectrum is available in closed form, $\rho_m=(L/\beta_m)^4$, with $\beta_m$ the roots of $\cos\beta\cosh\beta=1$. Under a bounded-support square-integrable propagation kernel the tangent operator is Hilbert-Schmidt, so the infinite-dimensional capacity is a well-defined Fredholm-determinant supremum. The optimal signaling law is a dual-budget generalized water-filling with one Lagrange multiplier for curvature power and one for phase excursion. From the exact nonlinear phase-only aperture law we derive the coherent tangent channel with an explicit Frechet remainder bound and a multi-chart atlas for large excursions. At the receiver, curvature inferred from noisy phase samples by second differences has a pentadiagonal noise covariance with spectral norm $\Theta(\Delta x^{-4})$. A deterministic diagnostic suite measures each mechanism against its closed form, with a null run beside every claim: the computed spectrum matches $(L/\beta_m)^4$ to relative error $3.78\times10^{-15}$; the tangent remainder has fitted slope 1.0000; the dual-budget law is solved with both multipliers strictly active to a KKT residual of $3.90\times10^{-15}$; the derivative-noise bound is approached to 0.999981; and the differentiation-limited branch is observed at exponent 0.2175, then collapses to 0.0576 once the mode count saturates at the Shannon number, while a flat-propagation null run holds at 0.2019. Curvature is thus a well-posed, gauge-invariant coordinate, and below the Shannon number the gauge, rather than the medium, governs the scaling.

How news media frame and emotionally code a public health emergency shapes public risk perception and trust, yet outbreak-coverage dynamics remain understudied for low- and middle-income countries (LMICs). We examine sentiment and stance in English-language Bangladeshi coverage of the 2026 measles outbreak – the country's most severe in two decades, with over 97,000 suspected cases and 600 deaths across 61 of 64 districts, unfolding after the 2024 change of government and a 2024-2025 vaccine stockout. Using the Internet Archive, we build a reproducible corpus of 403 headlines from seven national outlets (396 in-window in 2026), label them for binary sentiment and four-way stance via a large language model under a locked codebook, and validate against a two-coder human-adjudicated gold standard (n=153; Cohen's kappa=0.89 stance, 0.75 sentiment). Aligned to the DGHS epidemic curve, coverage grew significantly more negative (56% to 88% negative; Cochran-Armitage z=4.12, p<.001) and risk-amplification framing intensified (44% to 84%; z=4.15, p<.001). Media negativity lagged incidence, tracking cumulative mortality. Contrary to the political backdrop, blame remained a minority frame (~9% overall) and was overwhelmingly systemic (32 of 37, 86%) rather than directed at named actors. The pipeline offers a scalable, transparent method for LMIC outbreak-media analysis; Bangladeshi coverage amplified threat far more than it assigned political blame.

Distributed data management systems deployed in untrusted environments rely on Byzantine Fault-Tolerant (BFT) consensus protocols to tolerate malicious failures. DAG-based BFT protocols improve throughput by letting validators disseminate transactions concurrently and by scaling execution across multiple workers. However, imbalances in workload or resource capacity can still degrade performance significantly. This paper presents Flamingo, a load-balancing protocol for certified DAG-based BFT protocols that addresses imbalance at both the ordering and execution layers. At the ordering layer, Flamingo periodically migrates client accounts away from overloaded validators, adapting to skewed submissions and heterogeneous validator capacity while preserving correctness under Byzantine faults, with migrations taking effect only through the committed log. At the execution layer, Flamingo redistributes committed transactions across executor workers using a deterministic, order-preserving scheduler that balances load and minimizes cross-worker data movement, without centralized coordination or costly distributed commit. Built on top of Narwhal and Tusk, our prototype shows that Flamingo recovers throughput and latency under workload skew, validator heterogeneity, and shifting hotspots, adds negligible overhead when the system is balanced, and needs load balancing in both layers, since resolving only one shifts the bottleneck to the other.

As mobile robots become increasingly integrated into everyday environments, privacy risks arising from onboard cameras have become a growing concern. Ultra-low-resolution (ULR) RGB can mitigate visual privacy exposure at the source, but ULR appearance alone substantially limits semantic and spatial understanding. We therefore introduce a privacy-preserving asymmetric sensing setting that combines high-resolution (HR) depth with ULR RGB, preserving dense geometry while restricting fine-grained visual information. To address the severe information imbalance between HR depth and ULR RGB, we propose a joint 2D framework using HR geometry to guide semantic-oriented RGB reconstruction and RGB-D segmentation. Despite reliable frame-level predictions, consistent scene-level understanding remains challenging under the asymmetric HR depth–ULR RGB setting. We therefore develop an end-to-end 2D-to-3D pipeline that consolidates 2D semantic features for 3D segmentation. Experiments on ScanNet show that our method achieves the best 2D and 3D segmentation performance among privacy-preserving approaches and delivers the strongest zero-shot transfer to SUN RGB-D and SceneNN. Privacy recoverability analysis shows that our proposed HR depth–ULR RGB input reduces the recoverability of sensitive data, and real-robot experiments demonstrate the utility of the resulting 3D semantics for object-goal navigation.

Microscaling quantization techniques are increasingly used to represent neural network parameters with 8 bits or fewer while preserving near-full precision accuracy. However, applying these methods efficiently in convolutional layers is not straightforward. A naive approach transfers full-precision weights and activations to processing units and quantizes each tensor twice, resulting in much more memory movement than expected. Additional overhead comes from the activation tensors, whose sizes grow substantially because of the im2col transformation applied before quantization. We propose MicroQonv, a way to combine microscaling with convolutional layers' forward and backward operations by quantizing each tensor only once and quantizing the activation tensor before applying a modified version of im2col: channel-batch-first im2col. MicroQonv reduces the quantization cost by a factor of $\times2$ for weights and gradients, and by up to $\times9$ for activations, at a negligible accuracy cost. It reduces memory movement and storage by up to $\times7.53$ compared to their full-precision counterparts. This way, MicroQonv reduces microscaling-quantized activation memory movement by $\times3.5$ for state-of-the-art object detection models YOLOV8nano and $\times2.2$ for YOLOV26nano. It also enables 4-bit microscaling in a quantized latent replay strategy for continual learning at the edge, improving accuracy by +5.7% to +11%.

This paper presents a high-fidelity direct numerical simulation (DNS)-fluid-structure interaction (FSI) framework for rigid-liquid-flexible spacecraft dynamics under microgravity conditions. The liquid-gas flow is simulated with the incompressible two-phase solver implemented in DIVA, validated against FLUIDICS experiments conducted aboard the International Space Station (ISS). The flexible appendages are described by a rotating assumed-mode plate model that accounts for geometric stiffening. The fluid and structural operators are coupled through a Dirichlet-Neumann fixed-point algorithm with Aitken relaxation, and a closed-system mechanical energy balance is used as an a posteriori diagnostic to assess the energy imbalance of the partitioned discretisation. The coupling strategy is validated against an experimental free-decay sloshing benchmark, and its numerical consistency is assessed through spatial sensitivity studies of the energy-balance defect. Prescribed-motion, rigid open-loop, and flexible open-loop simulations of a spin-up manoeuvre are compared to isolate the effect of structural feedback on the sloshing response. Reduced liquid models identified from the different simulation architectures exhibit different predictive capabilities when embedded in the same rigid-flexible plant. A controller synthesized from the reduced model identified from the flexible simulation is replayed in the nonlinear CFD-FSI environment. The reduced model reproduces the principal attitude and actuator responses for the considered manoeuvre but does not recover the detailed nonlinear sloshing-load history. The framework provides a high-fidelity environment for analysing coupled spacecraft dynamics, identifying control-oriented models, and assessing reduced-model-based control strategies beyond their linear design representation.

Continuous-aperture and holographic MIMO systems motivate signaling that operates directly on large electromagnetic apertures rather than on a few antenna ports. This paper is the empirical companion to the operator-theoretic curvature-domain framework: a reproducible benchmark of curvature-domain signaling under a common scalar aperture-channel model, stress-testing the theory's dual-budget generalized water-filling law against the modal bases used in near-field and holographic MIMO. All methods share the same apertures, quadrature, Fresnel or Green-function propagation, phase-only constraints, power, phase-energy and curvature-energy budgets, receiver noise, phase quantization and training assumptions. The compared coordinates are curvature-regularized eigenmodes, raw phase coefficients, Fourier phase modes, polynomial and Zernike-like wavefront modes, near-field matched-focus profiles, random and optimized RIS phase codebooks, and SVD water-filling upper bounds. The claim is deliberately limited: curvature-domain signaling is a gauge-invariant, physically realizable coordinate system that can approach the phase-space SVD water-filling bound with fewer stable modes where derivative noise, phase quantization, sampling density or ill conditioning limit conventional bases. It is not a claim of new electromagnetic physics, nor that curvature modes dominate every baseline. We prove the SVD upper-bound relation for the discretized phase-control tangent space, derive pairwise-error and perturbation bounds, and report capacity, retained modes, symbol error, robustness, quantization, sampling, regularization, conditioning and cost, with 95% bootstrap confidence intervals on Monte Carlo results. A verdict table locates the regimes where curvature-domain signaling is engineering-relevant: moderate-to-high phase noise, coarse phase quantization, and non-Fourier-diagonal channels.

This study highlights the distribution of Standard Arabic (SA; H(igh) variety) and Colloquial Arabic (CA; L(ow) variety) across X and Facebook. 16754 public posts were collected via Python, with 10000 retained as the net dataset. Posts were classified into 7 discourse categories: *politics, technology, science, business, culture, fun,* and *sports*. Bivariate analyses, including Chi-square tests and Cramer's V (CV), examined associations among platform, discourse category, and diglossic choice, while binary logistic regression with Platform x Discourse Category interactions tested whether these associations varied across platforms. Findings reveal that there are significant associations between discourse category and diglossic choice on X, chi-square(6, *N* = 5000) = 600.35, p < .001, CV = .347, and Facebook, chi-square(6, N = 5000) = 1249.52, p < .001, CV = .500. Across platforms, platform was also associated with diglossic choice, chi-square(1, N = 10000) = 262.16, p < .001, CV = .162. Binary logistic regression further shows higher odds of SA use on X than Facebook in the political reference category (*OR* = 1.31, p = .0028), with significant platform-by-domain interactions for Culture (OR = 2.65), Fun (*OR* = 6.34), Sports (*OR* = 26.71), Science (OR = 0.41), and Technology (OR = 0.71). The study concludes that the diglossic use of SA and CA contributes to the growing body of research on digital discourse, unveiling that the digital age reshapes but does not erode diglossic boundaries, giving rise instead to a reconfigured digital diglossia.

Quantum computing is moving from research laboratories towards early commercialization and broader socio-technical adoption, supported by sustained hardware progress and a rapidly expanding open-source software ecosystem. This momentum is especially visible on GitHub, where many quantum and hybrid software projects coexist around frameworks such as Qiskit, Cirq, PennyLane and Amazon Braket. However, this ecosystem remains fragmented, making it difficult to understand who shapes quantum software, where expertise is concentrated, how collaboration flows across organizations and disciplines, and which actors connect otherwise separated communities. This paper presents Entangle, a data-driven analysis of the open-source quantum computing ecosystem on GitHub. Starting from 71 domain keywords, Entangle identifies more than 1,500 quantum repositories, 27,000 contributors and 400 organizations, revealing an ecosystem strongly organized around four leading industrial vendors, but also supported by 2,387 contributors who connect projects, organizations and domains. These findings provide practical evidence for responsible quantum innovation by making visible patterns of influence, dependency, collaboration and knowledge transfer. They also offer actionable indicators for strategic decisions on investment, hiring, partnerships, ecosystem stewardship and capacity building. More broadly, Entangle shows how open-source intelligence can support a more transparent, measurable and governable quantum software ecosystem, helping align technical development with responsible innovation, public–private coordination and long-term sustainability.

It is well known that the choice of basis functions in hp-FEM heavily influences the stability and the computational cost in order to obtain an approximate solution. For simplicial elements in two and three space dimensions, tensor-product-like basis functions built from Jacobi polynomials with different weights yield optimal properties with respect to condition number and sparsity. In this paper we construct such a high order basis for the Nédélec spaces and for the Raviart-Thomas and Brezzi-Douglas-Marini spaces, modifying existing definitions of Zaglmayr so that the functions belong to the Nédélec spaces of first kind and the Raviart-Thomas spaces. The bases are designed so that the Nédélec space of second kind and the BDM space are extensions of the respective other spaces. In the second part of the paper we introduce biorthogonal basis functions for $H(\text{div}, {\Omega})$ continuing previous research for $H^1({\Omega})$ and $H(\text{curl}, {\Omega})$. These functions can be expressed in closed form as sums of tensor products of Jacobi polynomials, which allows for fast computation of the $L^2$-projection.

Audio-language models (ALMs) integrate acoustic perception with the knowledge encoded in language models, enabling contextual understanding of auditory events. Making these capabilities practical on devices with limited memory and computation motivates our focus on small ALMs with fewer than 200M parameters. We introduce a recipe that brings together architecture, data, and three-stage training to build Mizar, a 159.3M-parameter ALM. Its architecture connects a compact CED-Small audio encoder to SmolLM2-135M through a frequency-merging mapper. With supervision drawn from ReasonAQA, AudioMCQ, and AVQA, the model undergoes three training stages: audio-language alignment (Stage 1), audio-dependent fine-tuning (Stage 2), and post-training (Stage 3) aimed at strengthening weak skills while retaining learned capabilities. Across five random seeds, Mizar achieves mean accuracies of 52.92% on MMAU, 42.42% on MMAR, and 36.02% on ADQA-clean, surpassing the previous best-performing ALM below 200M parameters on all three benchmarks. It also supports local inference on a single CPU: on questions from the MMAU benchmark, the mean latency from opening the audio file to generating a complete answer is 1.09 seconds. Code and checkpoints are available at https://github.com/KaiyangLi1992/Mizar_159M.

Removing an object from a real image requires more than synthesizing plausible content within a mask: the method must suppress residual object features, preserve the unedited scene, and generate replacement content that is consistent with the surrounding background. This paper approaches object removal from a stage-based perspective and proposes a zero-shot framework for constrained latent inpainting with a frozen pretrained Stable Diffusion model, requiring no task-specific training or model fine-tuning. The method integrates SAM-based mask construction, BLIP image-caption conditioning, DDIM inversion, background-weighted masked null-text optimization, decoder self-attention masking, hard outside-mask latent anchoring, and localized renoise–denoise refinement into a unified pipeline. The method is evaluated through qualitative examples, quantitative local-consistency metrics, and ablation studies. The results demonstrate effective object removal and context-consistent replacement content. The ablations indicate that background-weighted masked NTI is particularly beneficial for structurally complex backgrounds, whereas the no-NTI variant is sufficient in other evaluated examples. Repeated refinement further reduces object remnants and boundary artifacts remaining after the primary editing pass.

Pretrained generative Diffusion Transformers (DiTs) capture rich pixel-level visual and language-conditioned structure through large-scale image and video generation training. A growing line of robot policies builds on this generative prior, but how it should be transferred to control remains unclear, and existing approaches commonly instantiate this transfer through future visual prediction. We ask a more basic question: what a pretrained generative DiT actually contributes to action learning, and how this prior should be adapted for control. We introduce NowWAM, a future-target-free co-training formulation that denoises the current observation and predicts robot actions from the same visual stream, directly coupling the native generative objective to the action-facing representation across the denoising trajectory. Under matched controlled settings, past and future visual targets perform comparably, while restricting training to the clean endpoint substantially reduces robustness, suggesting that a separate future target is not essential for generative adaptation, while the denoising trajectory remains an effective interface for control. On LIBERO-Plus, NowWAM reaches 87.7% with FLUX2-Klein, improving over the future-target co-training baseline by 6.1 points while halving training visual tokens (784 to 392) and reducing step time from 2.85 s to 1.63 s, a 1.8x speedup. With the pure text-to-image Z-Image backbone, NowWAM further reaches 87.8%, showing that strong control adaptation is not tied to video generation or image-editing backbones.

Local geometric mixing localizes geometric mixing by requiring geometric convergence to equilibrium in total variation only over finitely many transitions. It accommodates local convergence rates and captures rapid local equilibration, even when global mixing is much slower. We establish and discuss local geometric mixing bounds through Dobrushin contraction. We then apply this approach to Diffusion Path Monte Carlo, a recently proposed Markov chain Monte Carlo method, aimed at leveraging advances in score-based modeling, whose ideal transitions coincide with those of the Proximal Sampler. Our analysis covers both the ideal method and its implementable Metropolis-adjusted counterpart, providing mixing guarantees under minimal assumptions. For the ideal method, these guarantees complement recent spectral gap estimates, which we develop into mixing time bounds.

Claims about AI safety reach audiences well beyond the AI community, yet many rely on opaque evidence or static assessments, when supporting evidence is accessible at all. We present the Systemic Risk Index, an open evaluation pipeline and dashboard built to make empirical evidence more transparent and traceable to the public. Our work organizes 19 public benchmarks into four systemic-risk categories defined by the EU GPAI Code of Practice—CBRN, cyber offense, harmful manipulation, and loss of control—and evaluates models using harm-preserving perturbations and simulated deployment contexts. The interactive dashboard lets users alternate between average and worst-case aggregation, vary how model capability affects the aggregate score, and trace each risk rating to its benchmark evidence. Across 18 models, scores fall by 14 to 37 points under worst-case aggregation, highlighting information that can be hidden by an average assessment of model risk. LLM judges show agreement with human graders comparable to human–human agreement ($\kappa = 0.78\text{--}0.82$), and a blind audit finds that $83\%$ of sampled transformations preserve the original harm. In a survey ($N = 21$), most participants report that scores are easy to understand and that the dashboard encouraged them to view model evaluations under different settings

We study online fair division, where indivisible goods arrive one at a time and must be allocated immediately and irrevocably among $n\ge2$ agents. Against an oblivious adversary that fixes the input in advance, we ask what fairness guarantees randomized online algorithms can achieve. We present the Deficit Allocation (DA) algorithm, which improves the previously known high-probability approximation to PROP1 from inverse-logarithmic to inverse-doubly-logarithmic in $n/\delta$, where $\delta$ is the failure probability. DA does not require $\delta$ as input. With probability at least $1-\delta$, a refined variant achieves an arbitrarily close approximation to proportionality when no single item is too valuable relative to an agent's proportional share; without any restriction on item values, it obtains an arbitrarily close approximation to PROP$k$ for $k$ doubly logarithmic in $n/\delta$. For $n2^{-2n}\le\delta\le1/2$, our bounds on the approximation factor and $k$ are asymptotically tight for the stronger fractional benchmark that divides each item equally among the agents who value it positively. We also prove that algorithms that use only which agents value each item positively, but not by how much, cannot improve upon the logarithmic dependence of uniformly random allocation. However, other fairness notions admit weaker guarantees. For EF1, every randomized online algorithm has an input on which any positive approximation succeeds with probability at most $e^{-\Omega(n)}$. For EFX, this probability is at most $1/n!$ with only $n+1$ items, tight within a factor of $(n+1)/2$; for MMS, it can be at most $5/6$. Finally, allocating each item uniformly among the agents who value it positively gives, with probability at least $1-\delta$, an arbitrarily close approximation to EF$k$ for $k$ logarithmic in $n/\delta$, and this dependence is necessary for this rule.

Accurate localization of the bronchoscope within the bronchial tree is essential for clinicians to be able to reach target lesions, perform biopsies and avoid misidentification of airway segments during diagnostic and therapeutic procedures. However, existing navigation systems typically rely on patient-specific CT scans or additional external sensors, increasing cost, setup time and patient radiation exposure. This work presents BronchoTop, a real-time, RGB-only framework for topological bronchoscopy localization that eliminates the need for patient-specific data. BronchoTop estimates scope location relative to a generic airway model through four modules: lumen detection and tracking, lumen-branch label association, probabilistic scope location estimation, and switch verification. By using only standard bronchoscopy video input, BronchoTop provides practical, real-time navigational assistance to physicians. Evaluation on phantom, simulated and real data demonstrates state-of-the-art accuracy, improving existing approaches performance by over 20% on real bronchoscopy sequences. BronchoTop is the first published framework including both the localization algorithms as well as all the real data used, together with code to generate additional simulations, encouraging and facilitating further developments and benchmarking. The results highlight BronchoTop's potential to enhance procedural safety, efficiency and accessibility in clinical and robotic bronchoscopy.

We present LightMIS, a scalable family of ultra-lightweight convolutional networks for 2D binary medical image segmentation without a learned stage-wise decoder. LightMIS aligns the outputs of a five-level encoder to a common resolution using Scale-Aligned Projection blocks, aggregates them once, and refines the fused representation with an Adaptive Fusion Cascade. The cascade combines Adaptive Kernel Fusion with the proposed Progressive Receptive Fusion module, which uses temporary channel expansion, complementary depthwise receptive fields, and progressive cross-branch information transfer. We evaluate LightMIS-T, LightMIS-S, and LightMIS using five-fold cross-validation under a common nnU-Net v2.3.1 protocol on DRIVE, Kvasir-SEG, DSB18, BUSI, ISIC-2017, and ISIC-2018. Full LightMIS contains 0.131 M parameters and requires 0.575 GFLOPs for a $3\times256\times256$ input, achieving modality-macro Dice and IoU scores of 86.71% and 78.99%, respectively. Mobile U-ViT obtains 86.75% Dice and 79.07% IoU, so the observed differences are 0.04 and 0.08 percentage points. Relative to Mobile U-ViT, nnWNet, and nnU-Net, LightMIS reduces parameter count by 90.58$-$99.61% and GFLOPs by 82.54$-$96.14%. On an Arm Mali-G52 MC2 GPU, all LightMIS variants achieve full GPU delegation, with median delegated latency ranging from 53.31 ms for LightMIS-T to 138.31 ms for LightMIS. These results demonstrate a favorable accuracy-complexity trade-off and on-device execution feasibility for the evaluated tasks. The code is publicly available at https://github.com/AndreiiArhire/LightMIS.

Model Predictive Path Integral (MPPI) control relies on stochastic trajectory sampling, and its performance under limited rollout budgets depends strongly on the structure of the proposal distribution. Conventional Gaussian perturbations, even when temporally correlated, transmit high-frequency jitter to the executed commands when only a few rollouts are available. We introduce Spike-MPPI, a motoneuron-inspired proposal in which a common stochastic drive recruits pools of motor units whose discharges are filtered by twitch kernels, yielding sparse, temporally coherent perturbations. We evaluate Spike-MPPI on legged locomotion and manipulation tasks with a Franka FR3 arm in simulation, as well as on a physical robot. Spike consistently produces the smoothest executed commands across all tasks, including against spectrum-matched Gaussian proposals, and a planar ablation identifies the common drive as the largest contributor to this smoothness among the tested motor-pool mechanisms. In manipulation, this smoothness is also accompanied by improved task performance and success, such as in ball balancing, where Gaussian, pink-noise, and spectrum-matched proposals largely fail. On hardware, Spike also yields the shortest task completion times and the smoothest commands. These results show that, beyond its power spectrum, the statistical structure of the sampling is a key design factor for finite-budget MPPI control.