Papers

A robot that assists people must often act on what a person has shown rather than said: which of several identical cartons was pointed at, or which box was handled. The plan is executed from the final scene, whereas the evidence occurs earlier, possibly on objects that have since moved. We propose instance-anchored interaction evidence (IAE), which registers every object of the final scene to its public identifier, keeps each identity through the video by backward mask propagation, and describes every frame by the geometry between hands, forearms and these instances. An evidence network trained only from task outcomes scores the instances. For pointing tasks, a grammar-constrained dynamic program trained with a structured loss decodes object-destination programs; symbolic programs handle reference disambiguation and, without learning, episodic tasks. On 1,255 WatchAct benchmark requests, scored by symbolic execution, IAE reaches 64.2% plan success on implicit-intent tasks against 27.5% for a 32B vision-language model (strict success 49.7% against 15.4%), and 46.4% against 27.0% on restoration, reversal and imitation without task-specific training. Controls with the same perception overlays, the same 32 frames, forward tracking, or a relation model trained on the same labels do not explain the gain. Given IAE's evidence as text with its meaning explained, the same language model reaches 57.4%: most of the gain comes from the instance-anchored evidence, and the explicit programs add 6.8 points at a fraction of the cost. Pointing remains the hardest case, with 16.9% strict success. The code is available at https://github.com/WeiZhou96/iae-watchact.

To stream long videos while maintaining visual quality and temporal consistency, Self Forcing mitigates exposure bias through self-rollout training on self-generated histories with key-value (KV) caching. To keep memory manageable, it detaches historical caches, preserving forward dependencies between chunks but severing the backward gradient paths. We introduce Connected Self Forcing, a training framework that reconnects gradient paths across autoregressive chunks, allowing feedback from later predictions to guide how earlier context is generated. These connections go beyond historical KV-writing: gradients pass through generated latents into the computations that produced them, linking the generation of earlier context to its use in later predictions. To make this connected training memory-efficient, we develop shortcut gradient replay, which recovers cross-chunk gradients without retaining the full rollout computation graph. Integrated with distribution matching distillation, Connected Self Forcing trains historical chunks according to both their direct supervision and their contribution to subsequent generation. Experiments on autoregressive video generation show improvements in long-horizon visual quality and temporal consistency, without changing the inference procedure.

We investigate turbulent dynamics from large to sub-proton scales, the kinetic wave activity and resulting proton heating in three-dimensional hybrid-kinetic simulations of freely decaying balanced and imbalanced Alfvénic turbulence. We considered initial fluctuations with same amplitude and spectral shape but different degrees of correlation between velocity and magnetic field perturbations (i.e., normalized cross-helicity), which define the corresponding turbulent regimes. We found that our balanced turbulence simulation exhibits stronger energy dissipation and more efficient proton heating compared to the imbalanced case. On the large fluid scales, the two simulations undergo distinct nonlinear dynamic and energy cascades, leading to different spectral properties once turbulence is well developed. At sub-proton scales, fluctuations are predominantly oblique with right-hand polarization, a property common to both regimes. In contrast, only the imbalanced simulation displays substantial wave power parallel to the guide magnetic field, with clear signatures of ion-cyclotron waves. These waves are excited in regions with strong temperature anisotropy that arise during the initial nonlinear steepening of Alfvénic fluctuations, a process that simultaneously generates parallel-propagating proton beams. Furthermore, non-Maxwellian features emerge in velocity space, indicating different heating mechanisms in balanced and imbalanced regimes and providing evidence for the role of kinetic instabilities in regulating the turbulent dynamics. Overall, these results show that cross-helicity is a key parameter controlling the large-scale evolution, kinetic activity, and energy dissipation in collisionless turbulent plasmas.

This paper investigates distribution network reconfiguration under demand uncertainty using a two-stage stochastic formulation. The network topology is selected in the first stage, whereas the electrical variables are determined separately for each scenario. The deterministic equivalent is formulated as a MISOCP problem that accounts for power balances, losses, voltage constraints, and radiality. Experiments on 12 networks, ranging from 17 to 10,561 nodes, consider three demand scenarios. Reconfiguration reduced expected losses in all instances, with reductions ranging from 3.23% to 65.23% and averaging 31.76%, while also reducing the number of voltage violations in the ex post evaluation by approximately 55%. Eleven of the 12 problems were solved to optimality within the prescribed time limit. However, the additional benefit of stochastic modeling was negligible because of the homothetic structure of the scenarios considered, which preserves the spatial distribution of loads. The results show that spatially heterogeneous scenarios are important for the representation of uncertainty to meaningfully affect reconfiguration decisions.

We study episodic constrained Markov decision processes with adversarial losses under stochastic hard constraints. Specifically, starting from a known strictly feasible policy with margin $d$, we seek to obtain optimal regret while satisfying the expected cost constraints in every episode. In this setting, Stradi et al. (2025) show that a carefully designed mixing rule attains regret of order $\widetilde{\mathcal{O}}(\sqrt{T}/\min\{d,d^2\})$. Interestingly, they also provide a lower bound of order $\Omega(\sqrt{T}/\rho)$ for the same setting, where $\rho$ is the Slater margin of the offline problem and can be much larger than $d$. In this work, we build on their approach to obtain optimal regret dependence on these margins. Specifically, we propose MA-OPS, an algorithm that combines an optimistic search for the Slater margin with a pessimistic evaluation of the selected policies to safely learn a policy with a large feasibility margin. This policy is then used to minimize regret while satisfying the constraints at every episode. In particular, we show that MA-OPS attains regret $\widetilde{\mathcal{O}}(\sqrt{T}/\rho + 1/(d\rho))$. Finally, we provide a matching lower bound, showing that the dependence on $T$, $d$, $\rho$ in the regret bound is optimal up to logarithmic factors.

Geodynamic simulations involve solving the Stokes equations, where boundary conditions play a crucial role in governing internal dynamics and surface topography evolution. A common approach is to model the interface in contact with air as a free surface. This interface is stress-free and deforms in response to internal force balances, which is essential for understanding processes ranging from rifted margins to subduction. However, a free surface can introduce instabilities and oscillations in the numerical solution, commonly known as the "drunken sailor" instability, forcing the use of small time steps. Geodynamic models therefore often employ stabilization terms based on the gravitational effect of free-surface motion. To date, these terms have been derived primarily on a heuristic basis. Here, we apply variational calculus to the energy-minimization formulation of incompressible Stokes flow. Using the Gateaux derivative, we derive the complete weak form for a deforming free-surface domain and show that the stabilization terms arise naturally from the free-surface assumption. We consider the linear-viscosity case and identify two boundary-integral contributions: one associated with internal viscous deformation and the other with gravitational body force. Two benchmark cases demonstrate the consistency of the derived formulation with previously proposed stabilization approaches. Our results provide a mathematical foundation for the commonly used gravity stabilization and identify an additional viscous boundary contribution.

Machine learning interatomic potentials (MLIPs) are transforming atomistic simulations by accessing unprecedented length and time scales. While pretrained equivariant graph neural networks achieve robust zero-shot performance for near-equilibrium properties across broad chemical spaces, their translation to complex materials dynamics remains fundamentally challenged by out-of-distribution reactive states, representation biases, and computational scaling limits. In this Review, we examine how physics-driven specialization extends the applicability of MLIPs to complex dynamical systems. We systematically evaluate the structural trade-offs in MLIP design: the undersampling of highly strained configurations, the prohibitive computational overhead of high-order message-passing architectures, and the necessity of nonlocal interactions for open and field-coupled systems. Through four demanding application contexts - electrified interfaces, compositionally fluctuating open systems, multiphase evolution, and large-scale fracture - we establish a framework for observable-specific validation. Highlighting the complementary roles of universal foundation models and task-specific potentials, we emphasize that targeted adaptations must be rigorously benchmarked against intended observables. We conclude with a roadmap for developing physically consistent, hardware-aware force fields that seamlessly connect electronic-structure accuracy to macroscopic materials phenomena.

In many engineering design problems, the objective and constraints depend on the state: the solution of a PDE determined by the design parameters. We consider improving a design while holding selected state observables near trusted values, which we call state preservation constraints. Constrained Bayesian optimisation handles these with a learnt feasibility model, but struggles with this problem's highly anisotropic feasible set. Our central idea is to pre-compute the set of controls whose linearised constraint response stays within tolerance, thereby pulling back the state-space constraint into design space. This linearisation defines an ellipsoid from which we can efficiently draw a large number of well-spread candidates. The underlying linear response map is refined online, and the ellipsoid is rebuilt accordingly. We demonstrate the method end-to-end on our key application - Tokamak divertor optimisation under plasma-boundary preservation.

N-version object detection (OD) is an approach to diversifying detection results using multiple models or input frames and reducing detection errors by aggregating individual results. Diversity and consistency across multiple detection results are critical information for characterizing the reliability of possible configurations of N-version OD systems. However, existing performance metrics such as mAP and Accuracy fail to capture these factors, as they are defined solely on the final outcome after aggregation. To overcome this limitation, we propose two reliability metrics particularly defined for N-version OD, namely coverage of errors in OD (Cov_OD) and certainty of accurate prediction in OD (Cer_OD), which can be computed from individual detection results without relying on voting strategies. We empirically demonstrate the unique features of the proposed metrics through a case study of N-version OD for a vehicle in an autonomous-driving simulator. We show that the proposed metrics can be used for 1) selecting version combinations with complementary error characteristics, 2) choosing an effective voting strategy based on diversity and consistency profiles, and 3) guiding incremental construction of N-version OD systems via pairwise two-version analysis. These results highlight the importance of the metrics that can guide the design of reliable N-version OD applications.

Quantum machine learning, and quantum neural networks (QNNs) in particular, are advancing fields with growing potential. Although systematic comparisons of QNN configurations have been explored primarily for classification tasks, comparatively little attention has been given to regression problems, particularly financial time series forecasting. This study presents a large-scale systematic comparative evaluation of QNN component configurations for financial time series forecasting, using the GBP/USD spot exchange rate as a case study. A grid search across encoding methods, ansatz designs, qubit counts, layer depths, and cost functions yields 1,368 distinct model configurations, each evaluated in terms of prediction accuracy, computational cost, and convergence behaviour. The results reveal unique insights into how the choice of methods influences performance, such as that gate selection and arrangement are more critical to model success than raw parameter count, and that entanglement is a system-level property of the full circuit rather than solely at the ansatz level. The best-performing QNN configuration achieves an $R^2$ score of 0.985, outperforming a classical BiLSTM baseline. Additionally, the impact of real quantum hardware noise is assessed through execution on the IQM Emerald device, revealing that gate errors and decoherence represent a significant barrier to practical deployment, with gate selection and circuit depth identified as key determinants of hardware noise resilience. Overall, the findings provide practical architectural guidance for QNN design and establish a baseline characterisation of QNN noise sensitivity on near-term quantum devices.

False negatives remain a critical limitation of computer-aided diagnosis (CAD) systems for breast cancer screening due to delayed detection and treatment. To address this issue, we propose a counterfactual data augmentation strategy that generates healthy mammograms by "erasing" lesions from anomalous images, thereby enriching the training distribution. We train a Denoising Diffusion Probabilistic Model on BI-RADS 1 (healthy) mammograms and use a RePaint-based sampling strategy to inpaint realistic normal tissue within annotated lesion bounding boxes. The resulting healthy counterfactuals replace annotated lesion regions with realistic healthy tissue while preserving patient-specific anatomical structure, as supported by similarity metrics between real and generated images. Image realism was further assessed by radiologists and found to be consistent with the original dataset quality. We evaluate counterfactual augmentation across four representative classifier architectures: a convolutional neural network (ConvNeXt), a vision transformer (ViT), a vision-language model pre-trained on mammogram-report pairs (Mammo-CLIP) and a multi-scale attention-based multiple-instance learning framework (FPN-MIL). Experiments conducted on the VinDr-Mammo dataset show improvements in sensitivity across all architectures, particularly at 80% fixed specificity, contributing towards more reliable CAD systems for breast cancer. Code is available at: https://github.com/ines03garcia/diffusion-based-counterfactual-generation.

Ultrafast science is rapidly evolving from a collection of specialized laser experiments into a facility-scale, data-intensive research endeavour. High-repetition-rate femtosecond and attosecond sources, synchronized multidimensional diagnostics, and increasingly nonlinear laser-matter interactions now generate heterogeneous data streams whose volume, speed, and complexity often exceed the capacity of conventional manual workflows. This chapter examines how machine learning can become an enabling layer for next-generation ultrafast facilities, with emphasis on the Extreme Light Infrastructure (ELI) ecosystem. We discuss the scientific drivers for artificial-intelligence-assisted experimentation, propose a facility-scale digital architecture spanning acquisition, ingestion, storage, analytics, machine learning, adaptive feedback, and user interfaces, and review representative examples from ELI ALPS, ELI Beamlines, and ELI-NP. The central thesis is that machine learning should not be treated merely as an offline data-analysis tool; rather, when constrained by physical insight and embedded in robust data infrastructure, it can support real-time diagnostics, surrogate modeling, anomaly detection, experimental optimization, and progressively autonomous discovery workflows.

Chirality-Induced Spin Selectivity (CISS) refers to structural chirality governing spin transport in chiral systems, with experimental signatures reported in both transport experiments and photoemission. Beyond influencing transport, recent experiments indicate that adsorbed chiral molecules can influence the magnetic state of adjacent ferromagnetic layers, suggesting the existence of chirality-induced effects whose origin and quantitative characterization remain lacking. Here, we investigate chirality-induced long-range magnetic interactions in a hybrid molecular/ferromagnetic multilayer beyond the conventional direct transport coupling. We demonstrate the long-range nature by decoupling the adsorbed chiral molecules from the magnet using an insulating spacer. We show that self-assembled monolayers of chiral polypeptides with a helical structure can act on a ferromagnetic multilayer via an enantiomer-dependent effective magnetic field that may be mediated by chiral phonons through the insulating layer. Using Magneto-Optical Kerr Effect (MOKE) microscopy with on-chip reference regions free of molecules, we directly quantify this field by measuring precisely the reproducible shifts of the local hysteresis loops. The extracted effective field reverses sign for opposite enantiomers and correlates with the enantiomeric excess. These results provide a direct and quantitative link between molecular chirality and magnetic response in hybrid heterostructures, establishing an experimentally accessible benchmark for a CISS-related effect.

Tokenizer choice affects multilingual language modeling, but vocabulary capacity is finite and vocabulary size is often constrained: improving representation for some languages often comes at the expense of others. We therefore ask whether tokenizer choice matters equally across languages, a question that the current literature leave unanswered. To this end, we train 123 language models spanning 54 tokenizers. In the main comparison, architecture, training corpus, training-token budget, and optimization are held fixed, so the models differ only in their tokenizer. We find that tokenizer choice matters more for languages with less language-model training data: across the 54 tokenizers, the standard deviation of a language's bits-per-byte (BPB) increases as its model training-data share decreases (Spearman rho = -0.52 over the 31 trained languages and -0.69 over the 28 written with word boundaries). Leaving a language out of tokenizer training raises its BPB in every language we study, and the penalty tends to be larger for languages with less language-model training data. Giving lower-resource languages a larger share of tokenizer-training data, however, does not unconditionally help those languages: both equal weighting and an allocation inverting the shares with respect to the language model training data increase their BPB, particularly when language-model training repeats data. Finally, which intrinsic tokenizer properties are associated with better BPB differs across languages, providing further evidence that what makes a good tokenizer depends on the language. We find that the metrics quantifying these properties can be successfully used to predict downstream models' pairwise BPB rankings, suggesting a practical strategy for screening tokenizer candidates before training language models.

We report the discovery of a new radio pulsar PSR J0103$-$72 in the Small Magellanic Cloud (SMC) through a reanalysis of Parkes/Murriyang archival observations targeting supernova remnants and pulsar-wind nebulae in the SMC. Previously two pulsars were discovered in these observations, and PSR J0103$-$72 is the third which we independently detected in two observations of its target field. It has a spin period of 1.327 sec and a dispersion measure (DM) of $272\pm13~\mathrm{pc\,cm^{-3}}$, making it the slowest known radio pulsar in the SMC with the second-highest DM. We did not identify any plausible supernova remnant or pulsar-wind nebula associated with the pulsar. Subtracting the predicted Galactic foreground leaves a DM contribution of $\sim 230$–$242~\mathrm{pc\,cm^{-3}}$, which may suggest a long path through the ionized medium of the SMC, enhanced ionized gas along the sightline, or both. Although these data had previously been searched using both FFT and fast folding algorithms, our FFT-based reprocessing found PSR J0103$-$72 through search techniques which have previously been highly effective in discovering a large number of pulsars in other surveys. This discovery shows that candidates with weak Fourier-domain significance can reveal faint, narrow-duty cycle pulsars missed in earlier extragalactic pulsar searches.

Electromagnetic spectrum monitoring increasingly requires flexible analysis beyond task-specific recognition and detection. Multimodal large language models offer a unified interface, but extending vision-language models (VLMs) to raw I/Q signals requires tokenization that balances fidelity against a strict budget. For signals, dense encoding causes token costs to grow with observation length, whereas fixed-resolution compression may discard short-duration or localized signal evidence. Thus, we propose BATok, a budget-adaptive signal tokenizer that adjusts token capacity to the input length while allocating that capacity according to the signal content. BATok constructs candidate representations from signal-derived features using lightweight multi-resolution branches, then combines a local energy prior with learnable queries to resample these representations into compact signal tokens. The number of tokens adapts to the input length while remaining strictly bounded. The resulting tokens are projected into the language embedding space of VLMs. We further introduce EMSpec-Instruct, a multimodal instruction dataset aligning raw I/Q signals, waterfall images, and language supervision for modulation recognition, structured detection, and language-conditioned signal grounding. Experiments show that BATok learns effective signal representations and achieves competitive performance across all tasks.

Standard convolutional neural networks for turbulence super-resolution are restricted to uniform structured grids, preventing their direct application to the non-uniform and non-orthogonal meshes commonly used in wall-bounded flows and complex geometries. We introduce a finite-volume convolution unit (FVCU) that removes this restriction. The key idea is to construct convolutional layers using finite-volume operators that are applicable to non-uniform and unstructured meshes. To this end, we use an exact representation of a 3D convolution as a linear combination of successive axis-wise convolutions. We adopt axis-wise gradient, averaging, and identity operators, as they can be evaluated on arbitrary polyhedral cells using finite-volume discretization. The resulting FVCU is well-defined on general meshes, serving as a drop-in replacement for conventional convolution layers. The FVCU is embedded in a super-resolution neural network and trained to reconstruct 3D velocity fields from $4\times$ coarsened low-resolution fields in turbulent channel flow on a non-uniform mesh, and in circular-pipe flow on a non-orthogonal O-grid mesh. Inspired by the role of the Van Driest damping function in near-wall subgrid-scale modelling in large-eddy simulations, this function is supplied as a conditional input to encode wall proximity. An ablation study confirms that the network interprets this input in a physically consistent manner. In both configurations, the model accurately reconstructs instantaneous velocity fields, Reynolds-stresses and their anisotropy, and energy spectra beyond the Nyquist limit of the low-resolution mesh. On an elliptic pipe unseen during training, the model reproduces the main flow features and local statistics, with a modest overprediction of small-scale energy. These results demonstrate a framework for extending CNN-based super-resolution to general finite-volume grids.

Growing automation makes collaborative robots work in more variable environments, increasing the need for adaptation. We propose a human-in-the-loop online training framework combining a digital twin (DT), reinforcement learning (RL), and human demonstrations. Unlike DTs used mainly to generate synthetic data before task execution, our DT is synchronized with the physical system in real time through camera feeds, allowing the virtual robot to update its observations and policy from real-world feedback. A dual actor framework integrates imitation learning (IL) without adding a direct imitation loss to the RL actor, so demonstrations can guide adaptation instead of manual reprogramming. The proposed framework is demonstrated on the Ufactory Xarm5 collaborative robot, where the robot's end-effector aims to reach the target position while avoiding obstacles. The experiments show that the framework can resume training after a change in the physical workspace and that, with a fixed set of non-optimal demonstrations, the dual actor framework achieves a much higher final success rate than two methods that add an imitation loss to the actor. The same pattern holds with real human demonstrations collected in virtual reality (VR): with demonstrations that never reach the goal, the dual actor framework reached 83-100% mean deterministic evaluation success, against 0-17% for the two imitation-loss methods.

The integration of inline Artificial Intelligence (AI) models into critical network infrastructure is fundamentally bottlenecked by the high latency and synchronization overhead of CPU-mediated packet processing. While legacy GPU offload and recent CPU-bypass frameworks attempt to bridge this gap, they remain trapped in proprietary ecosystems or still rely on the host CPU and coarse-grained batching to coordinate stateful telemetry and AI pipeline execution. In this paper, we propose AGP, a novel framework that promotes the GPU from a passive accelerator to a primary data-path controller, answering what changes architecturally when the GPU autonomously owns the complete packet-to-inference pipeline. By enabling GPU-native packet processing, contention-free in-GPU stateful aggregation, and a persistent mega-kernel for continuous AI inference, AGP removes the CPU from the critical path. Our evaluation demonstrates that native in-GPU packet processing sustains line-rate throughput while being over 6.1x more power-efficient. Furthermore, our integrated Intrusion Detection System (IDS) stress test eliminates the legacy CPU-GPU synchronization tax, reducing end-to-end latency by 7.9x (up to 35x p99) and accelerating whole-system inference throughput by 9.3x using only ~2% of the GPU's thread capacity.

We consider a once-excited random walk on $\mathbb{Z}^d$, $d \ge 2$, where the walk on its first visit has a bias of constant strength $\beta>0$ toward the origin and moves like a simple symmetric random walk on subsequent visits to that site. We show that the walk is recurrent and prove an almost sure spherical shape theorem: the trace after $n$ steps is asymptotically a Euclidean ball centered at the origin, with radius proportional to $n^{1/(d+1)}$, and the local times have an asymptotically conical profile. This result confirms a conjecture of Kozma (2007) on the shape of the trace for this model.

Unauthorized model distillation, in which a model is trained on the outputs of a proprietary large language model (LLM), is a growing threat to model providers. We study distillation inference: determining whether a suspect model was distilled from another model or trained independently. We formulate this problem as a hypothesis test and estimate the behavior expected under each hypothesis by training shadow models: distilled shadow models learn from the teacher's reasoning traces, whereas independent shadow models learn only from reference answers. The auditor measures how closely each model predicts the teacher's reasoning outputs and then uses the shadow models to convert the suspect's score into a calibrated p-value. In a preliminary study using Qwen2.5-7B as the teacher and Llama-3.2-3B for the suspects, our test achieves a true positive rate of 1.0 at a significance level of 0.02. These results demonstrate the feasibility of using distillation inference to detect distillation attacks.

High harmonic generation (HHG) has emerged as a powerful probe of ultrafast electron dynamics in solids. However, in strongly correlated materials, where electronic excitations arise from many-body interactions rather than single-particle band dispersion, the field remains in its infancy. Here, we report HHG up to the 23$^{rd}$ order in the prototypical Mott insulator La$_2$CuO$_4$. We observe a robust 4 eV cutoff that is independent of driving field strength, revealing an intrinsic energy scale defined by the Hubbard bands. Using time-resolved high harmonic spectroscopy, we disentangle the distinct contributions of carrier creation, intraband acceleration, and interband recombination to the nonequilibrium radiation spectrum. We demonstrate that photo-doping triggers a fundamental reorganization of the many-body manifolds above a critical excitation density, a process that renormalizes the band curvature and suppresses nonlinear intraband currents, thereby directly imprinting the signature of many-body band reconstruction onto the harmonic emission. These results establish HHG as a sensitive probe of subtle correlation-driven electronic reconstructions and highlight that many-body interactions enable extreme nonlinear optical responses that are highly tunable in nonequilibrium states.

Learning to tackle long-horizon, goal-conditioned tasks requires an agent to reason over extended timescales and act across a broad range of states. In principle, Hierarchical Reinforcement Learning (HRL) addresses both challenges through the interaction between action (temporal) and state (spatial) abstraction. First, using an action abstraction to represent temporally extended behaviour as options reduces the effective decision horizon. Second, enabling different state abstractions at each level of the decision process permits greater data aggregation for learning. However, realising these two benefits of a hierarchical policy depends on learning an appropriate action abstraction. Current HRL algorithms fail in one of two ways. Some discard distinctions between options needed for optimal control, undermining hierarchy altogether. Others retain unnecessary distinctions, preserving horizon reduction, but forfeiting coarser state abstraction. In this work, we characterise three desiderata for an action abstraction. We introduce Q-Shaped Options (QSO) to address all three. QSO builds on an architecture with distinct state-value functions, Q functions and policies at each level of the hierarchy. It learns the action abstraction between consecutive levels as a shared encoder shaped by their respective Q functions. The low-level Q function uses the option as a goal, encouraging the abstraction to retain distinctions necessary for optimal control. The high-level Q function uses it as an action, encouraging unnecessary distinctions to be discarded. Across offline goal-conditioned locomotion and manipulation environments, QSO learns semantically meaningful option spaces and outperforms baselines, achieving non-zero performance in tasks where all other evaluated algorithms fail.

Knee osteoarthritis (KOA) progression prediction can support patient monitoring, requiring the integration of multimodal data and multidomain expertise. Moreover, isolated risk estimates provide limited insight underlying a prediction. To automate the progression assessment workflow and reduce manual effort while providing interpretable findings and supporting evidence, we present OA-MAP, an autonomous multi-agent framework for evidence-grounded assessment of structural and pain progression in KOA. The system incorporates modality-specific agents including MRI, X-ray, and clinical agents, together with a coordinator agent. This framework can autonomously recruit specialist agents, select tools for prediction and analysis, and retrieve literature as external evidence based on user request and available patient information. An uncertainty-informed human-in-the-loop mechanism enables clinicians to review and correct intermediate findings, triggering recomputation of affected results. We evaluate the prediction models using 600 participants from the FNIH Osteoarthritis Biomarkers Consortium cohort. On the test set of 100 participants, the fusion models achieve AUROCs of 0.80 for structural progression and 0.68 for pain progression. A case study illustrates how OA-MAP combines risk estimates with intermediate findings, cross-modal conflicts, literature support, and uncertainty indicators to support interactive review.

Many key-value (KV) caches are compressed before anyone knows what will be asked of them: a document cached for retrieval, a prompt prefix shared across requests, the memory of a long conversation. The prevailing approach scores KV entries by rehearsal: the model rereads the context and keeps the entries it attends to, assuming that the more completely a cache rehearses its context, the better it remembers it. We show that under tight budgets this assumption backfires: rehearse everything, remember nothing. At a 3% keep ratio, rereading the whole context keeps 31.5 of 96.5 points on RULER, and on LongBench's natural-text tasks it falls below methods that rehearse nothing at all. The cause is that a cache keeps what it rehearses: rereading spreads the budget across the whole context, so the answer's own entries survive at little more than chance. Like a student before an exam, a cache remembers more by testing itself than by rereading. Two principles follow: rehearse what will be read, and rehearse as much as there is. We instantiate them as Attic-KV (Attic for short), a training-free rehearsal in which the model quizzes itself with question-answer pairs that quote the context, alongside anchor tokens in a content-adaptive amount. Changing only the rehearsal lifts three hosts that score it in three different ways: Attic alone is the best training-free method in all eight settings we test on RULER and LongBench's natural-text tasks, and plugged into the gradient-based KVgrad and the trained RestoreKV+, it raises them by up to 17.1 and 28.1 points. Its advantage grows as the budget shrinks, reaching 41.9 points over full rereading at a 3% keep ratio, and on long documents it compresses faster than rereading them.

Ion correlations set the structure of polymer electrolytes and depend on the host permittivity, which is low and varies between polymers. A functional that carries these correlations from molecular simulations to larger inhomogeneous systems must transfer across domain sizes, geometries and coupling strengths. We introduce a neural density functional for the ions of a salt-doped polymer melt with built-in integrability, continuous spatial symmetry, the Noether identities, perfect screening in stable bulk states, and parameters that refer to no grid or box. The permittivity enters only through the Bjerrum length, an explicit argument constrained by an exact relation: the free-energy derivative with respect to the Bjerrum length is the Coulomb correlation energy. Trained by force matching on planar profiles at five Bjerrum lengths spanning 8-30 monomer diameters, the functional predicts, without retraining, density profiles in fourfold longer boxes and under two-dimensional fields, bulk structure factors and the long-wavelength number response, also at a concentration and a permittivity left out of training. It reproduces this energy identity on Coulomb energies that never enter the training. Its nonlinear density dependence captures correlations that a quadratic pair closure misses: the number response grows several-fold with concentration, and at low concentration the coupling lowers it through ion association, below the ideal value at the strongest coupling, while the host sets it in the concentrated melt. At fixed coupling, it is more accurate than three neural architectures and loses little accuracy with a quarter of the training data.

Let $A(X)$ denote the free Abelian topological group over a Tychonoff space $X$. We prove that if $A(X)$ and $A(Y)$ are topologically isomorphic and $X$ is countably compact, then $Y$ is also countably compact. Thus countable compactness is an $A$-invariant and, consequently, an $M$-invariant; this answers Open Problem~7.10.1 of Arhangel'skii and Tkachenko.

The Simplex method is among the most widely used approaches for solving linear programs. Starting at a vertex solution of the feasible region, the algorithm proceeds through a sequence of basis exchanges (known as pivots), each corresponding to a move along an improving edge of the polyhedron toward a better vertex. A central challenge affecting the efficiency of the Simplex method is degeneracy, which can cause long sequences of pivot operations that fail to change the current vertex solution. In this paper, we prove the existence of a pivot rule which is able to escape degeneracy and follow any given shadow edge-direction when initialized at some compatible basis, with a linear number of degenerate Simplex pivots. As a byproduct of our result, we obtain an improved bound on the number of degenerate Simplex pivots needed to solve linear programs defined on 0/1 polytopes.

A large body of research measures model coherence based on output variance without adequately considering competing causes. We identify two such causes, ambiguity and indifference, and we introduce a set of 175 questions where contradicting answers cannot easily be explained by either. We then measure incoherence in terms of contradictions when resampling answers to the same question. In contrast to other methods our metric has high specificity, and only ranks models as incoherent when the issues are glaring. Even so, we find narrow finetunes score poorly. Inspecting inconsistencies flagged by our method, we find that model organisms from the literature display severe issues such as identity conflation, introspection failures and rationalizations. These findings suggest that the pathologies induced by narrow finetuning may limit what these models can tell us about coherent misaligned behaviour.

Hyperon polarization in relativistic heavy-ion collisions sensitively probes thermal vorticity and the operator structure of local equilibrium. We investigate an isotropic generalized uncertainty principle (GUP) via a spin-independent scalar deformation of the one-particle phase-space density of states in the local rest frame (LRF). We employ the Belinfante-Rosenfeld local-equilibrium operator, the standard on-shell dispersion relation, and the Pauli-Lubanski/axial-Wigner representation of the spin-$1/2$ observable, retaining terms linear in the GUP coefficient and in hydrodynamic gradients. The same scalar GUP factor appears in the scalar and axial phase-space contributions and cancels from the local mean-spin ratio at fixed spacetime point and momentum. Thus, the leading GUP contribution to an experimentally relevant polarization observable arises through the Cooper-Frye freeze-out integrals and can be expressed as a covariance between the LRF momentum invariant and the local spin kernel. This cancellation is a consequence of the scalar density-of-states modification within a fixed pseudogauge and does not imply pseudogauge independence of the polarization observable. For $\Lambda$ hyperons, a Maxwell-Boltzmann thermal estimate shows that the dependence on freeze-out temperature and fluid-dynamical correlations is substantially larger than the Fermi-Dirac correction, which remains negligible over the temperature range considered. Present LHC precision corresponds only to a formal linear sensitivity of order $10$-$30~\mathrm{GeV}^{-2}$. Since this interval lies beyond the domain in which the linear GUP expansion is reliable, no exclusion limit is inferred. Differential measurements of hyperon polarization as functions of momentum, rapidity, collision centrality, and azimuthal harmonic structure provide the most direct experimental avenue for testing the predicted covariance dependence.