Papers

A medical model's benchmark score does not establish that the same conclusion holds under a different evaluation. This study tests whether claims about model ranking, score reliability and screening performance survive changes in cohort, prompt, negative spectrum, specified prevalence and operating threshold. We audit three medical vision-language models (BioMedCLIP, CheXficient, and MedSigLIP) and a general-domain OpenCLIP comparator on 12,200 chest radiograph records from four datasets (Montgomery, Shenzhen, TBX11K, and VinDr-CXR). Five fixed prompt families yield 244,000 model–image–prompt scores. No model leads every cohort and reliability criterion. Prompt-family changes alter AUROC in 21 of 48 multiplicity-controlled comparisons. Replacing healthy controls with sick non-tuberculosis controls reduces AUROC by 0.075–0.306 across all four models. On VinDr-CXR, the three medical models distinguish tuberculosis from no-finding controls substantially better than from pneumonia or lung tumor; their AUROC point estimates for both named diseases fall below 0.5. CheXficient has documented VinDr-CXR pretraining exposure, which limits the interpretation of its results. Thresholds chosen for 95% sensitivity on TBX11K training retain that constraint by point estimate in only four of sixteen target evaluations. A five-seed supervised source model reaches 0.999 AUROC on TBX11K validation but 0.629 on each of two external cohorts. Conservative exclusion of perceptual-overlap candidates narrows this gap without closing it. These retrospective, single-task results show that discrimination, score reliability and threshold retention support different portability claims. Evidence for chest X-ray tuberculosis screening should identify the complete evaluation specification rather than attribute clinical portability to a checkpoint alone.

Simulation based inference (SBI) circumvents the challenge of intractable likelihoods by using a simulator that generates data given parameter values. For instance, neural likelihood estimation (NLE) estimates the likelihood function by training a neural network to perform conditional density estimation on simulated data given corresponding parameters. However such density estimation is only feasible for relatively low dimensional data. We extend the scalability of SBI methods to a higher dimensional problem: long sequences with a complex dependency structure. We introduce Neural Composite Likelihood Estimation (NCLE). This divides the sequence into smaller, equal-sized batches. Instead of training NLE to estimate the likelihood for an entire sequence, we estimate the likelihood for each batch separately. The product of these forms an approximate composite likelihood (CL), and we perform frequentist inference using methods from the CL literature: we get a point estimate from maximising the approximate CL and obtain confidence intervals by estimating the Godambe information matrix. We demonstrate the effectiveness of NCLE with experiments on time series models.

We present CRISP (Contact-RIch Simulation Platform), a high-fidelity physics engine tailored for complex multi-contact simulations such as tight-tolerance robotic manipulation. Achieving high physical fidelity in robotic simulation requires both expressive modeling of geometry and contact interactions, as well as accurate numerical resolution via robust collision detection and contact solvers. However, existing simulators often either rely on limited support for geometric representations and simplified modeling of contact interactions, or employ numerical resolution methods whose accuracy or robustness is inherently constrained. Accordingly, we develop a new simulator that supports diverse geometric representations with accurate optimization-based collision detection, and combines contact modeling with robust augmented Lagrangian-based contact solvers. This integration enables efficient and consistent detection of contact information across complex geometries while accurately resolving multi-contact constraints without problematic relaxations, which is essential for simulating contact-intensive and sharp interactions. We validate the physical fidelity of our simulator against state-of-the-art platforms and further demonstrate its capabilities through complex robotic manipulation scenarios. CRISP is publicly available at https://github.com/INRoL/crisp.

In this paper, we prove that a mean system utilization smaller than one is a necessary condition for the feasibility of real-time systems. Such systems are defined as stable. Stable systems have two distinct states: a transient state, followed by a steady-state where the same distribution of response times is repeated infinitely for each task. We prove that the Liu and Layland theorem holds for stable probabilistic real-time systems with implicit deadlines, we provide an analytical approximation of response times for each of those two states and a bound of the instant when a real-time system becomes steady.

We present the first complete machine learning method for accelerating plane-wave density functional theory (DFT) in materials under the projector augmented wave (PAW) formalism. We formalize seven criteria that a Complete Neural Electronic Initializer must satisfy for practical end-to-end PAW DFT acceleration. Applying these to prior work reveals two structure-dependent components, augmentation occupancies and spin initialization, whose absence prevents existing acceleration methods from providing complete reference-free initialization. We show that omitting these components can eliminate or reverse the acceleration obtained via models that only predict the smooth valence density. We satisfy the missing requirements by introducing AugNet, a general equivariant model for PAW augmentation occupancies, and the first general spin density model for materials, which predicts the smooth spin-difference density and spin-difference PAW augmentation occupancies using predicted magnetic moments to constrain the global magnetic state. Combined with existing valence density models, our full method satisfies all seven criteria and forms a fully reference-free electronic initializer for materials DFT, requiring no electronic quantities from a converged target calculation. We show that perfect initialization could cut PAW DFT wall time by 40-52%, and our method recovers up to 62% of this saving, reducing end-to-end DFT wall time by up to ~25% on unseen structures while preserving converged energies.

Topology optimization (TO) represents a significant step towards automating the design process: given a working simulation, TO can produce a viable prototype at the press of a button by differentiating the simulation and iteratively improving the design. In practice, however, TO is riddled with “magic numbers”—hyperparameters whose tuning significantly affects the outcome. Finding the right values typically requires not only deep problem-specific knowledge but also extensive trial-and-error. While practitioners can use surrogate-assisted hyperparameter optimization as an alternative, this approach requires strictly limiting the number of hyperparameters through careful problem formulation. Here, we propose differentiating TO itself using automatic differentiation. This yields “hypergradients” that allow us to tune these hyperparameters in tandem with the primary optimization. We show that evaluating just one or two steps of TO is sufficiently informative and that the method scales favorably to thousands of hyperparameters at an expense comparable to only a few standard TO runs. We demonstrate this approach on stress-constrained and compliance problems, with the latter utilizing a neural parameterization of the density field.

We study macroscopic thermalization in the ferromagnetic Ising model with a weak generic (highly nonlocal and many-body) random quantum perturbation. We prove that a single typical perturbation makes every initial state $|\Phi(0)\rangle$ in a specified energy shell thermalize: a measurement of a large class of macroscopic observables in the time-evolved state $|\Phi(t)\rangle=e^{-i\widetilde H_Lt}|\Phi(0)\rangle$ yields their thermal equilibrium values, within a prescribed precision and with overwhelming probability, at sufficiently large, typical times $t$. The result covers the full finite-temperature range, including the ordered phase, where we use plus boundary conditions. With periodic boundary conditions in the ordered phase, every energy eigenstate in the shell is shown to be an almost balanced macroscopic Schrödinger-cat state. Moreover, every pure initial state becomes such a cat at sufficiently large, typical times, with each branch in the corresponding macroscopic thermal equilibrium. The proof combines the theory of Roos, Sugimoto, Teufel, Tumulka, and Vogel with rigorous Ising large-deviation estimates.

Gender inequality remains a persistent structural feature of the labour market, shaping women's lifetime earnings and economic security. As artificial intelligence (AI) transforms organisational practices, there is growing concern that existing disparities may be unintentionally amplified through task automation, unequal access to upskilling opportunities, and differential returns obtained from technological change. In this paper, we examine how exposure to AI-driven innovation varies across male- and female-dominated occupations, with particular attention to differences across the skill and wage distribution. Using a novel dataset that links occupational characteristics to measures of AI exposure, we analyse how recent advances in Large Language Models (LLMs) and broader AI technologies are distributed across the labour market. Our findings show that, while AI exposure is generally concentrated in higher-skilled and higher-paid occupations for male-dominated occupations, female-dominated occupations display relatively uniform levels of exposure across both high-skilled, high-paid, and low-skilled, low-paid occupations. Moreover, we find that LLM-related exposure is higher in female-dominated occupations, while exposure to broader AI innovation remains more concentrated in male-dominated occupations. A triangulation of these results with existing literature suggests that women, particularly those in the most vulnerable positions (lower-skilled and lower-paid female-dominated occupations), may face greater exposure to forms of AI associated with task automation, job restructuring, reduction of wages and limited career progression.

Emergency department (ED) decision-making relies on heterogeneous clinical information, including patient history, vital signs, laboratory results, and electrocardiograms (ECGs). Vision–language models (VLMs) can jointly process these modalities, but strong predictive performance does not necessarily imply meaningful use of the correct patient's ECG. We term this failure mode ECG Mirage: apparent multimodal capability without useful dependence on patient-specific ECG information. We distinguish two forms: ECG neglect, where ECGs provide little predictive benefit, and ECG confusion, where matched ECGs outperform no-image inputs but not mismatched ECGs. To evaluate these behaviours, we compare predictions obtained with matched ECGs, outcome-discordant mismatched ECGs, and no-image inputs while holding the clinical text and prediction targets fixed. Across four VLMs on MDS-ED, matched ECGs provide no consistent advantage for either ICU admission or clinical deterioration prediction. We then train four restricted visual prompts using supervised learning followed by conditional direct preference optimisation, while keeping the VLM backbone frozen. The resulting models achieve balanced accuracies of 70.6% for ICU admission and 67.5% for deterioration and increase the matched-versus-mismatched performance gap to approximately 16.5 and 5.5 percentage points, respectively. Overall, our study identifies ECG Mirage in multimodal clinical prediction and introduces visual prompt tuning as an efficient mitigation strategy.

Deep learning-based visual odometry (VO) has achieved significant progress, yet most existing methods focus on a monocular approach, which suffers from scale ambiguity. Stereo VO provides real metric by its nature, but remains less studied in deep learning VO due to its high computational cost and modeling complexity. Recent advances in stereo matching and optical flow estimation have made dense visual correspondence increasingly accurate and reliable, but their complementary geometric information has not been fully exploited for VO. In this paper, we present SFVO, a correspondence-driven stereo VO framework that directly builds upon pretrained stereo matching and optical flow models. SFVO exploits pretrained stereo matching and optical flow models to estimate stereo and temporal correspondences. Instead of learning pose directly from images, SFVO maps learned correspondences into geometric constraints and predicts which points are trustworthy. To improve the reliability of visual correspondence-based geometric constraints, we introduce decoupled confidence maps for rotation and translation. This design better aligns the characteristics of visual correspondence and 6-DoF transformations. Extensive experiments on outdoor and indoor datasets demonstrate that SFVO achieves robust and accurate pose estimation with strong generalization capability. The code will be released.

Contact-rich manipulation requires policies to generate precise actions by reasoning over contact forces, robot configurations, and interaction histories beyond visual observations. Existing methods passively condition on force feedback rather than actively predicting future contact dynamics, limiting their ability to generate high-precision actions. To address this problem, we introduce Predictive Sensorimotor Representation (PSR) learning, a framework that learns a hierarchy of predictive representations from multimodal sensorimotor signals and integrates them into the action stream of a visuomotor policy. Specifically, during a pretraining stage, a multimodal Transformer is trained to learn a hierarchy of predictive representations by jointly forecasting future interaction dynamics. The learned hierarchy subsequently augments the action stream, enabling the resulting policy to exploit contact-relevant cues at multiple depths. We further instantiate PSR within a Vision-Language-Action (VLA) model, resulting in PSR-VLA, and evaluate it on six real-world contact-rich manipulation tasks. Experimental results show that PSR-VLA achieves 91.7% overall success, improving over $\pi_{0.5}$, ForceVLA-$\pi_{0.5}$, and ForceVLA2-$\pi_{0.5}$ by 30.0, 22.5, and 19.2 percentage points, respectively. These results demonstrate the effectiveness of the proposed PSR for force-aware, contact-rich manipulation. Videos of the tasks and stability tests are available at https://psr-vla.pages.dev/.

We study the dynamics of weighted backward shifts on Ces`aro spaces associated with leafless locally finite rooted trees. We first characterize their boundedness in terms of adjacent level cardinalities and edge weights. We then characterize their $\mathcal{F}$-transitivity by a growth condition involving level cardinalities, products of weights along paths, and a level-dependent Ces`aro factor. As consequences, we obtain criteria for hypercyclicity, weak mixing, topological ergodicity, and topological mixing. We also characterize the existence of nonzero orbit limit points and chaotic weighted shifts, the latter in terms of normalized fixed points and unit flows satisfying an explicit summability condition. Examples show that $\mathcal{F}_{\underline{d}>0}$-transitivity need not imply frequent hypercyclicity, that a nonhypercyclic weighted shift may nevertheless have a nonzero orbit limit point, and that topological mixing need not imply chaos.

Manipulating objects requires understanding not only their motion, but also the physical properties that determine it. For articulated objects, these include inertia, friction, and mechanisms such as springs or door closers, whose effects can vary with configuration and velocity. Such properties are not directly observable from appearance: visually identical doors may require very different effort to manipulate. Existing digital-twin pipelines recover primarily kinematics or assign static physical parameters from visual and language priors, which can yield physically implausible estimates. As a result, state-dependent mechanism dynamics remain unidentified and are not represented in standard asset formats. We present ForceTwin, a system for identifying physics-informed digital twins of articulated objects from instrumented human interaction. A person probes an object using a handheld force-sensing gripper, providing synchronized poses and interaction forces from which we estimate the articulation, parametric dynamics including inertia, Coulomb friction, viscous damping, and a structured neural residual capturing state-dependent mechanism forces. ForceTwin nearly halves the inertial-parameter error of a VLM prior. As a feedforward dynamics model for impedance control on a Spot and a Franka FR3, ForceTwin achieves 87% goal completion across nine object-embodiment pairs, compared with 60% using VLM-prior and 57% using kinematics-only twins, with the largest gains on objects whose strong mechanisms cause both baselines to stall. We further use the identified twins to train whole-body door-traversal policies and deploy them in the real world. Project Page: https://timengelbracht.github.io/forcetwin-website/

We compute the fully renormalized one-graviton-loop contribution to the self-mass of a massive scalar field in de Sitter space. The computation is performed using dimensional regularization and the divergences are absorbed by BPHZ counterterms. This self-mass can be used to study one-loop quantum-gravitational corrections to the dynamics of massive spectator scalars in inflation.

Skills can improve the performance of Large Language Model (LLM) agents by providing task-specific procedural guidance, while skill optimization further improves their effectiveness through iterative refinement. However, existing skill optimization methods typically represent skills as unstructured natural-language instructions, creating two key challenges: 1) Unstructured skills often lack explicit workflow-level guidance and contain substantial redundancy, making them difficult for LLMs to execute; 2) the vast search space of unconstrained natural-language skills makes skill optimization ineffective. To address these challenges, we propose representing skills as graph-structured natural-language artifacts. In graph-structured skills, each node represents an execution step together with its operational guidance, while directed edges encode context-dependent transitions between steps. Compared to unstructured skills, graph-structured skills can provide clear workflow-level guidance. Moreover, the proposed graph-structured skill can also facilitate skill optimization. Building on this structured representation, we introduce GraphSkillEvo, a population-based evolutionary optimization framework with mutation and crossover operators for graph-structured skills. By maintaining multiple candidate skills and combining effective components, GraphSkillEvo enables broader and more comprehensive exploration of the structured skill space than purely LLM-based iterative self-refinement. Extensive experiments across five agent benchmarks demonstrate that GraphSkillEvo consistently outperforms the strong skill optimization baseline SkillOpt, improving average accuracy by 4.01% on GPT-5.4-nano and 1.76% on GPT-5.4. Our code is available at https://github.com/ruisun7/GraphSkillEvo.

Behavioral failures can make a transformer appear to lack a world model even when it has learned faithful representations of its environment. We demonstrate this in TaxiGPT, a transformer trained on random walks through Manhattan whose failures have been interpreted as evidence of an incoherent internal map. Through mechanistic analysis and causal interventions, we show that the model represents intersections and streets, tracks its position, and uses a goal compass to navigate. We trace its failures to interference between superposed intersection features, which disrupts localization within the internal map. Affordance packing, which groups representations of intersections with the same legal moves, helps limit the consequences of these errors. Finally, we propose mechanistic indicators that we use to compare models and show that world-modeling capacities emerge at different stages of training. Our findings motivate a shift from asking whether a model has a world model to mechanistically studying its world modeling: the interacting capacities through which it represents its environment and uses those representations to guide behavior.

We develop single-loop stochastic projected damped extragradient methods for nonconvex–(strongly) concave minimax optimization, with complexity guarantees for both game stationarity (GS) and optimization stationarity (OS). Our approach combines a stochastic projected damped extragradient (SPDE) method with a recursive variance-reduced variant, VR-SPDE, both of which retain a single-loop structure. Under an unbiased stochastic gradient oracle with uniformly bounded variance, SPDE finds an $\varepsilon$-game-stationary point with stochastic first-order oracle (SFO) complexities of $O(\kappa\varepsilon^{-4})$ and $O(\varepsilon^{-5})$ in the nonconvex–strongly concave and nonconvex–concave settings, respectively, where $\kappa=L/\mu$. Under an additional mean-square Lipschitz condition on the stochastic gradients, VR-SPDE improves these GS complexities to $O(\kappa^{3/2}\varepsilon^{-3})$ and $O(\varepsilon^{-9/2})$, respectively. For an $\varepsilon$-optimization-stationary point, SPDE achieves SFO complexities of $O(\kappa\varepsilon^{-4})$ and $O(\varepsilon^{-6})$, while VR-SPDE achieves $O(\kappa^{3/2}\varepsilon^{-3})$ and $O(\varepsilon^{-6})$, in the two settings, respectively. These OS guarantees match the best-known bounds achieved by multi-loop methods while preserving a single-loop implementation. To the best of our knowledge, our results provide the best-known SFO complexity guarantees among single-loop stochastic first-order methods for the respective stationarity criteria and problem classes.

Two-dimensional magnetic semiconductors provide a unique platform for exploring the interplay between charge, lattice, spin, and light. The strong Coulomb interaction combined with long-range magnetic order enables efficient coupling between excitons and both lattice vibrations and spin excitations, thereby strongly influencing exciton dynamics and transport in these materials. Here, we develop a microscopic many-body theory of exciton–phonon and exciton–magnon interactions in mono- and bilayer CrSBr. We demonstrate that the quasi-one-dimensional character of excitons in CrSBr gives rise to strongly anisotropic exciton–phonon and exciton–magnon scattering, resulting in direction-dependent diffusion coefficients. Furthermore, we identify a distinct temperature dependence originating from one-phonon and two-magnon scattering processes, providing characteristic signatures of the underlying microscopic interactions. Our work establishes a microscopic framework for understanding exciton transport in magnetic van der Waals semiconductors and provides guidance for future optical spectroscopy and exciton transport experiments.

We study leptogenesis and Planck-scale remnant dark matter from primordial black hole (PBH) evaporation in a minimal Pati-Salam cosmology, with the gauge symmetry broken before inflation so that magnetic monopoles are diluted away. A singlet inflaton with a near-inflection potential enhances the curvature power spectrum on small scales, producing a narrow black hole population that briefly dominates the energy density. The Pati-Salam embedding ties the right-handed neutrino masses to the SU(2)_R breaking scale, and cosmological consistency forces both the scalar-sector Yukawa coupling and the heavy neutrino mass well below that scale. Two regimes emerge, depending on whether the black holes are hot enough to emit the lightest right-handed neutrino. Lighter PBHs drive non-thermal leptogenesis, while heavier ones require a thermal asymmetry and also affect it through entropy dilution. If quantum gravitational backreaction halts evaporation at the Planck scale, each PBH leaves a stable remnant whose present abundance scales as the inverse five-halves power of the initial mass. Suppressing remnants requires heavier black holes, but the lighter PBHs needed for non-thermal leptogenesis overproduce remnant dark matter. Thus, remnant dark matter and non-thermal leptogenesis are mutually exclusive. When the thermal contribution is included, a single initial PBH mass near 10^6 g accommodates the observed dark matter abundance, the heavy neutrino mass required by the baryon asymmetry, and the Pati-Salam breaking scale. The scenario predicts a stochastic gravitational wave background from Poisson fluctuations of the black hole distribution, within reach of future experiments, alongside high-frequency graviton emission constrained by future measurements of the effective number of relativistic species.

Robot assistance is particularly crucial in unfamiliar tasks, where users must understand task requirements while coordinating with the robot. Previous research offers mixed evidence on the role of robot proxemics in user engagement: some studies suggest closer proximity enhances interaction, while others report it can feel intrusive. In this work, we argue that perceptions of intrusiveness depend not only on proxemics but also on the frequency of robot interventions, and are strongly influenced by individual traits such as personality and demographics. We conducted an experiment with 32 participants who interacted with two assistive robots that provided similar task support but differed in their intervention strategies. Results indicate that overall engagement remains stable across conditions, yet affective responses and perceived intrusiveness vary significantly with personality traits. Moreover, personality shapes interaction dynamics differently depending on the robot's behavior. These findings emphasize that effective human-robot interaction should account for individual differences, tailoring robot behavior to maintain engagement while respecting each user's unique affective and behavioral profile.

Entropy minimization is a standard objective for test-time adaptation (TTA), but it can fail in imbalanced binary segmentation. Unlike image classification, dense segmentation aggregates thousands of pixel predictions, allowing the larger predicted class to dominate the update, pull minority predictions toward itself, and produce a degenerate mask as predictions saturate and their entropy gradients vanish. We theoretically establish this collapse in a shared-shift model. This analysis motivates Balanced-Anchor Prompt Adaptation (BAPA), which combines two complementary modules. The Class-Balanced Anchors (CBA) module selects high-confidence anchors separately from each predicted class and gives foreground and background equal total loss weight, preventing the larger region from dominating the update. Dynamic Prompt Adaptation (DPA) refreshes these anchors after each prediction update and optimizes only text-side prompt residuals while keeping the vision-language encoders frozen. This prompt-only update refines the foreground-background decision boundary without altering the pretrained dense visual representation. Across experiments from four domains, BAPA achieves the highest mean Dice among the evaluated methods. Factorized ablations further validate the complementary roles of CBA and DPA, supporting balanced prompt adaptation as an effective alternative to entropy minimization for test-time binary segmentation.

The core-collapse supernovae (CCSNe) 2022prr, 2023ucy, and 2024ljc exploded within the central regions of the nearby luminous infrared galaxy (LIRG) NGC6745. We present our follow-up data with the aim to characterise the spectrophotometric evolution of these CCSNe, determine their host-galaxy extinction, and to define their physical nature. The presented spectrophotometric data set covers the optical and near-infrared region for our three supernovae (SNe). The data are analysed and compared to other hydrogen-rich CCSNe, and a host-galaxy extinction value Av is derived based on broadband photometry for each of the targets. A spectral energy distribution (SED) for the host galaxy is constructed, and a combination of different galactic components are fitted to the SED to estimate the intrinsic CCSN rate and the age of the starburst of NGC6745. 2022prr is a SN2009ip-like event, whose light curve evolution consists of two consecutive luminous events A and B with peak magnitudes of Mr = -14.8 and -18.3 mag, respectively. The high degree of V and R-band polarisation (P=2.8%) near maximum light suggests interaction with highly aspherical circumstellar material. The spectral evolution of the SN is dominated by narrow Balmer emission lines, which have a complex evolution over time. 2022prr joins the growing sample of SN2009ip-like events that show early `flash ionisation' features of CIII, NIII, and HeII. 2023ucy is found to be a normal Type IIP SN. However, 2024ljc is a rapidly evolving Type IIb SN with a post-maximum r-band decline of roughly 3.2 mag by +41 d whereas normal IIb SNe decline typically by 1.5+/-0.3 mag during the same timescale. While the events exploded in a LIRG, the SNe are obscured by only moderate V-band host extinctions ranging from 0 to 1 mag. Based on the SED modelling we estimate an age of the host starburst to be 40Myr, and the intrinsic CCSN rate to be 0.4+/-0.1 SN/year.

Floquet engineering and cavity quantum electrodynamics represent two complementary descriptions of light-matter interaction, yet their precise connection beyond the small-coupling regime remains subtle. Here, by representing the cavity field in a large-photon-number phase basis, we show that the time-dependent Schrödinger equation of a classically driven system emerges directly from the fully quantized problem, without replacing the field by a classical trajectory or assuming a mean-field decoupling. This establishes a one-to-one correspondence between quasienergies and photon-sector-shifted quantum energies, while providing a direct quantum interpretation of the Floquet mean energy and Anandan phase. Because gauge invariance is exactly preserved at the level of the quantum Hamiltonian, the quantum-to-classical correspondence remains valid beyond perturbative and small-coupling regimes. We further show that light–matter entanglement is encoded in matter–harmonic correlations of the corresponding Floquet mode, thus identifying a semiclassical limit in which Floquet physics emerges without requiring the quantum field itself to approach a coherent classical state.

Latent world models enable planning by predicting the effects of actions in a learned representation space, but their predictions can become unreliable when test-time conditions differ from training. Existing test-time adaptation methods address this by updating parts of the pretrained model, often modifying millions of parameters and requiring a choice of which internal components to adapt. We introduce Sandwich-Residuals, a lightweight alternative that keeps the pretrained world model frozen and learns only small residual corrections around the predictor. The residuals are optimized online using the model's self-supervised prediction error and require no rewards, labels, or source-domain data. Across 21 conditions on the AdaJEPA benchmark, our method achieves $1.3\times$ the success rate of the frozen model while retaining 95% of the performance of the strongest AdaJEPA variant and adapting 97-99% fewer parameters. Under compound shifts, this advantage increases to $1.9\times$ the success rate of the frozen model, while remaining comparable to internal block adaptation. We further demonstrate the same adaptation principle on a DINO-WM model for 3-D manipulation. These results suggest that effective test-time adaptation of world models does not necessarily require modifying their pretrained internal weights.

Let $\gamma(G)$ denote the number of cliques in a graph $G$ and let $\Gamma(n):= \{\gamma(G):|V(G)|=n\}$ be the set of values of $\gamma(G)$ that can be attained on $n$ vertices. We improve on a result by Erdős and Erné to show that $| \Gamma(n)| \geq 2^{n-4\ln(2)\log^3(n)}$ for sufficiently large $n$.

Modern text-to-speech (TTS) systems are rarely deployed as unchanged pre-trained models. They are often adapted through supervised fine-tuning (SFT) or preference optimization such as DPO and GRPO. This raises a practical question for audio deepfake forensics: do fingerprints learned from a foundation generator remain valid after adaptation? We present GenTraceBench, a controlled benchmark spanning five TTS architectures, 16 pre-/post-training variants, and 49,728 utterances generated with fixed texts and speaker prompts. Under a train-on-foundation, test-on-adapted protocol, we evaluate binary detection, closed-set attribution, and open-set verification. DPO and GRPO generally preserve fingerprints, whereas some SFT and pre-training-data changes cause substantial drift; effect sizes vary across three forensic backbones. Repeated training runs confirm the largest W2V-BERT attribution drop, while a data-mixture control with comparable speech quality shows that composition change need not cause drift. In W2V-BERT verification, multi-shot enrollment reduces EER for the SFT condition from 44.4% to 11.0%, whereas the SingNet-only condition remains at or above 45% EER.

Let $m\geq1$, $k\geq0$, $n=m+k$, and write $x=(v,z)\in\mathbb R^m\times \mathbb R^k$, where $v$ is the active velocity variable and $z$ is passive with respect to $Y=v\cdot\nabla_y+\partial_t$. We consider \[ \operatorname{div}_x(A\nabla_xu)-Y(wu)=0, \] where $w=w(x)\in A_2(\mathbb R^n)$ and the measurable matrix $A$ has ellipticity and size controlled by $w$. For arbitrary such weights, including weights depending on the active variables, we prove a weighted hypoelliptic transfer estimate, a kinetic Sobolev inequality, local boundedness, the weak Harnack and Harnack inequalities, local Hölder continuity, and the strong minimum principle. The Harnack argument uses a full-variable weighted measure-data compactness theorem. An $\mathrm L^p$ velocity-averaging estimate gives compactness of smooth moments of $wu$, and weighted Poincaré inequalities in all diffusive variables reconstruct $u$. At a two-phase limit, testing with non-negative profiles produces transport directions given by weighted mean velocities. A $\mathrm{BV}$ argument applied to the limiting subsolution inequality yields the no-jump principle needed for expansion of positivity and a weighted kinetic ink-spots lemma. The theory includes the weight $|\lambda|^{1-2s}$ in the extension of Garofalo and Tralli for fractional powers of the Kolmogorov operator.

Achieving sufficient near-wall resolution remains challenging for wall-bounded turbulence simulations using weakly compressible smoothed particle hydrodynamics (WCSPH), as local particle refinement imposes restrictive time-step requirements. This work proposes a multi-scale near-wall approach within the Reynolds-averaged Navier–Stokes (RANS) framework to improve near-wall predictions without refining the SPH particle distribution. A local one-dimensional sublayer solver based on the simplified steady $k$–$\omega$ equations is coupled with each wall-adjacent fluid particle. A local flow-rate constraint and a friction-velocity-based iteration scheme are developed to close and solve the sublayer system, while a two-way shear-stress coupling with local feedback improves consistency between the two scales. The proposed approach is evaluated using turbulent straight-channel flows over a wide range of Reynolds numbers and a wavy-channel flow involving separation and recirculation. The results demonstrate improved near-wall velocity, turbulent kinetic energy and friction-coefficient predictions, satisfactory convergence, and good agreement with reference solutions. In particular, the approach improves the convergence of near-wall predictions where conventional wall treatments struggle to provide consistent results, and agrees closely with reference solutions at Reynolds numbers as high as $8.0 \times 10^{7}$ without near-wall particle refinement. For separated flow, convergence comparable to that of locally refined finite-volume simulations is obtained using an approximately uniform particle distribution. These improvements are achieved with limited computational overhead, supporting the application of particle-based methods to engineering flows involving wall-bounded turbulence.

Effective exploration in high-dimensional continuous control remains a central challenge in reinforcement learning. Planning-based methods address this by combining online planning with learned policies and value functions, but their components can become misaligned during training: learned sampling policies may diverge from planner behavior, while planning distributions stored in replay become stale as the model and value function evolve. Reanalysis can refresh these targets, but at substantial computational cost. We propose GEM-MPC, an MPPI-based reinforcement learning method that improves the interaction between planning and learning. GEM-MPC uses MPPI to combine a policy trained to clone the planner with a KL-regularized policy that explores around it, providing complementary exploitation and guided exploration within planning. We further introduce Gated Prior Distillation, which selectively learns from stored planning distributions only when they provide a better target than the current prior, reducing the impact of stale planning data without requiring full reanalysis. Across continuous-control benchmarks, GEM-MPC consistently outperforms existing planning-based baselines under lower computational budgets.

Designing effective Human-Robot Interaction in task-oriented settings requires carefully balancing user engagement with socially acceptable levels of robot intrusiveness. In this paper, we examine how different robot intervention strategies shape user experience, interaction dynamics, perceived intrusiveness, and sense of support. We compare two approaches: a continuous engagement-seeking robot strategy, and a context-aware strategy that selectively intervenes based on the user's state and task context. Both approaches rely on multimodal behavioral cues, including body orientation and attentional signals, to guide robot actions. We evaluate these strategies in a user study with 32 participants performing a task in a simulated hospital environment. Our findings show that higher interaction frequency does not necessarily lead to better engagement. Instead, we observe a systematic trade-off between perceived support and intrusiveness, influenced by factors such as physical proximity and user effort. These results provide empirical evidence that effective engagement in HRI depends on adaptive, context-sensitive intervention policies.