Papers

We study the $k$-Canadian Traveller Problem (k-ctp), an online navigation problem introduced by Papadimitriou and Yannakakis in 1991. Given an undirected weighted graph $G=(V,E,\omega)$ and two vertices $s,t\in V$, a traveller aims at navigating from $s$ to $t$ while minimizing the total distance travelled. However, up to $k$ edges of the graph may be blocked, and the traveller discovers that an edge is blocked only upon reaching one of its endpoints. The performance of an online algorithm for k-ctp is measured with the competitive ratio, which is the quotient between the distance actually traversed and the distance traversed by an omniscient traveller who knows all blockages in advance. A general lower bound $2k+1$ for the competitive ratio of deterministic algorithms remains valid even when the input graph is series-parallel, whereas constant competitive ratios can be achieved on restricted graph classes, such as cycles or unit-weighted outerplanar graphs. In this paper, we establish the first intermediate competitive ratio between constant and linear regimes. We prove that, on unit-weighted solid grid graphs, there exists an online algorithm with competitive ratio $O(\sqrt{k})$. This ratio is asymptotically tight and cannot be defeated, even on unit-weighted square grids. Finally, we show that both the unit-weighted and solidity assumptions are essential: going further with arbitrary edge weights (even on square grids) or non-solid unit-weighted grid structures yields a linear lower bound of $\Omega(k)$.

Generative Flow Networks (GFlowNets) learn policies that sample discrete objects in proportion to a positive reward. They are trained by minimizing balance objectives, whose exact satisfaction guarantees correct sampling. However, a comprehensive theory linking nonzero balance loss and approximate stationarity to sampling accuracy, especially under evolving sampling distributions and backward policies, is still lacking. To bridge this gap, we develop a unified theory for Flow Matching, Detailed Balance, and Trajectory Balance in this regime. We first bound terminal symmetric KL by the square root of the expected balance loss. Then, we show that approximate stationarity controls this distribution error up to a tangent-space residual measuring parameterization incompleteness. Finally, we establish finite-time convergence under fixed or controlled sampling and fixed, balance-trained, or separately updated backward policies. Under standard smoothness assumptions, with fixed or controlled sampling and a fixed or balance-trained backward policy, the expected loss is $O(\bar\epsilon^2+T^{-1})$ and expected terminal symmetric KL is $O(\bar\epsilon+T^{-1/2})$ under uniform coverage. For a separately updated backward policy, we characterize the sufficient update decay to approach the residual neighborhood and preserve these rates. These results connect balance objectives to sampling accuracy and clarify how parameterization, coverage, and moving components govern GFlowNet training.

Sequential social dilemma benchmarks capture diverse social interactions, but rarely test whether agents can infer hidden environmental rules through repeated interaction while adapting to other agents. We introduce OpenSocialJax, which extends Clean Up and Commons Harvest across an open-ended distribution of latent environment configurations, systematically varying spatial layouts and hidden rules governing tool crafting, payoffs, and resource dynamics. Each episode samples one configuration from tens of millions of possibilities and holds it fixed across five trials. We evaluate eight large language models, including three closed source models and five open source models. Our evaluation tests whether models can discover hidden rules and adapt to other agents in homogeneous and mixed teams. We find that rule discovery often improves performance but does not reliably sustain cooperation, as the cleaning burden becomes concentrated among fewer agents in OpenCleanup and every team eventually exhausts the shared resource in OpenHarvest. In our mixed arena, one cooperative agent was able to sustain the public good in OpenCleanup at a cost to its own return, but could not prevent resource depletion in OpenHarvest. Together, these results highlight the difficulty of adapting jointly to unfamiliar environmental dynamics and the changing behavior of other agents.

Parameterized Quantum Circuits have emerged as a central framework for variational quantum algorithms in the NISQ regime, with major applications in quantum simulation, combinatorial optimization, and, more recently, quantum machine learning (QML). By developing a new approach for computing dynamical Lie algebras associated with quantum circuits combining parametrized gates and non-parametrized entangling gates, we aim to extend the applicability of recent algebraic results to architectures that are actually used in quantum machine learning.

Long-context serving runs out of KV memory before the hardware reaches its useful throughput. HiSparse stores history in host memory and fetches only the selected entries, but it uses one cache budget at every load and does not handle draft-model speculative decoding such as DSpark and DFlash2. We present Adaptive DHiSparse, which supports speculative decoding over hierarchical sparse KV and adapts residency to load. It protects the required working set, retains additional history while GPU memory is spare, and recovers the rest through exact fused retrieval. DSpark and DFlash2 keep their proposal and acceptance rules: queries share missing-row retrieval, draft KV uses a window-bounded ring, and verification staging is reused across layer groups. At 100K context and concurrency 128, GLM-5.3 with DFlash2 delivers 793 output tokens/s per decode GPU, 1.45$\times$ HiSparse and 1.27$\times$ non-speculative adaptive serving. Held to the same latency bounds as HiSparse, it sustains up to 4$\times$ the concurrency and 2.3 to 3.7$\times$ the output, depending on the per-token bound. On DeepSeek-V4.1-Flash at 500K context and concurrency 512, Adaptive DHiSparse with DSpark delivers 1.85$\times$ the output of resident serving with DSpark. In a load step, optional residency reduces host fetches from 14.2% to 1.4% of selected rows at low load and releases memory under pressure.

We give an explicit example of a 9-dimensional 5-step nilpotent complex Lie algebra, generated by two elements, that is isomorphic to its complex conjugate but has no real form. In contrast to the earlier 10-dimensional examples, our proof is completely elementary and very short.

A camera-controlled video world model must keep a scene consistent when the camera returns to it. Memories that render stored features into the target frame can turn depth and pose errors into ghosting or duplicated structure. Positional transport avoids that resampling by relocating a historical patch's address rather than its content, but assigns full positional precision even to inaccurate projections. We present GLIMPSE, a global-local memory interface with band-wise multi-precision coordinates and geometry-conditioned spectral attenuation to reduce sensitivity to inaccurate projected addresses. Local patch memory supplies view-specific appearance, while recurrent global memory provides scene context. Footprint-guided retrieval selects historical patches under a fixed local read budget. Under matched retrieval and read budgets, GLIMPSE improves geometric correspondence and photometric consistency over deterministic positional transport.

We construct a linear map from the space of spin polynomials in the critical Ising model on a lattice to the state space of the $M(4,3)$ minimal model of conformal field theory. The kernel of the map consists precisely of null fields, i.e., spin polynomials that vanish in all correlations. For a polynomial $\Op$ which is not a null field, we define its scaling dimension $\Delta_{\Op}\in\Z_{\geq0}\cup(\Z_{\geq0}+\frac{1}{8}),$ to be the lowest $\Loper{0}+\Loperbar{0}$ eigenvalue in the decomposition of its image into homogeneous components. We show that as the mesh size $\delta$ goes to zero, the rescaled field $\delta^{-\Delta_{\Op}}\Op$ has a non-trivial scaling limit, when inserted in correlations with other such fields. Moreover, all these limits can be expressed in terms of correlations of primary fields of $M(4,3)$ and the Virasoro action in the continuum. The highest weight vectors of $M(4,3)$, the joint eigenvectors of $\Loper{0}$ and $\Loperbar{0}$ with the same eigenvalues $0,1/16,1/2$ for both, are the images of the constant $1$, the spin at a single lattice vertex, and the energy at a single lattice edge, respectively. For the proof, we construct a map from a larger lattice field space, accommodating also fermion and disorder insertions, to a direct sum of two fermionic Fock spaces. This direct sum contains contains the state spaces of both the $M(4,3)$ and the free fermion CFT. The main technical ingredient of the proof is a new version of the exact fermionic current action on the space of lattice local fields due to C. Hongler, K.K. and F. Viklund, combined with some new advances in discrete complex analysis. Our results are valid on arbitrary isoradial graphs.

Data deduplication relies on content-defined chunking (CDC) algorithms to decide where data is split. We observe that most CDC algorithms are modular: a base chunking algorithm fixes the byte-level cut-point condition, while high-level techniques (size normalization, min/max thresholds, caching) layer on top. Existing benchmarks conflate the two, obscuring which design choices actually drive performance. We isolate and evaluate eight state-of-the-art base chunking algorithms across four realistic workloads and four metrics: throughput, deduplication ratio, mean chunk size, and chunk-size variance. Further, we derive new formulas relating algorithm parameters to expected chunk size for AE, RAM, and MII. In doing so, we uncover a longstanding error in the widely-used AE parameterization reproduced across multiple papers and evaluation tools that produces chunks 13-39% smaller than intended and inflates AE's reported deduplication. Across our evaluation, AE is the most balanced base chunking algorithm across all four metrics, RAM's similarity to AE does not carry over to low-entropy data, and MII and PCI show limited advantages over simpler alternatives in storage-oriented workloads. We release all implementations, datasets, and analysis scripts as open source.

The fast pace in which protocols, frameworks, and technologies are being proposed for LLM-based agents results in a complex scenario for anyone aiming to create state-of-the-art AI solutions. Without an assessment on how different protocols, frameworks, and libraries work together to compose an Agentic Web solution, developers and organizations face different challenges (e.g., skilling, governance, infrastructure) even before becoming part of the Agentic Web ecosystem. In this context, we propose a system for performing Agentic Web evaluations called MASTIF (Multi-Agent System TestIng Framework) and present a baseline assessment of Agentic Web technologies, including 3 LLMs, 3 protocols, and 4 agent implementation frameworks, against 300 Mind2Web benchmark tasks. Results suggest that selecting a task-specific technology stack may result in gains of task completion rate, efficiency score, and social value stability scores; and reductions in verbosity and time to complete tasks. With the presented results, we expect to present a baseline in terms of metrics and results for this complex and fast evolving scenario for those aiming at designing, assessing, and developing multi-agent systems.

A sender with commitment provides additional information to an already informed receiver. Final posterior means must be more dispersed than initial means and less dispersed than the state, in convex order. We characterize when these two plausibility constraints are also sufficient, so that every distribution satisfying them can be induced. If the state has a positive density, they are sufficient exactly under interval-partition information: the receiver learns which interval of a partition contains the state and nothing more, where an interval may be a single point. Otherwise, a nonempty open set of distributions satisfying both constraints cannot be induced, even when the sender observes the receiver's information. We use the sufficient case to derive an optimal symmetric disclosure rule in a second-price auction where bidders initially know valuation intervals.

The feasibility and potential performance of optical clocks based on all-optical atom chips using neutral strontium atoms have been investigated. The key benefits of all-optical atom chips are their ability to trap alkali-earth atoms, and how easily these atoms can be manipulated and examined. The planar configuration of the chip enables ring photonic resonators to be incorporated into its structure, making it a promising candidate for active optical atomic clocks. Using the 'magic wavelengths' of the trapping optical fields (813 nm and 390 nm for strontium atoms) minimises frequency shifts in the intercombination clock transition caused by variations in optical field intensity. However, this frequency shift can be compensated for using optical AC Stark shifts, which are generated by the trapping optical fields and have frequencies below the 'magic frequencies'. Furthermore, it has been demonstrated that atom chip clocks can operate using solely repulsive, blue-detuned evanescent light fields when the attraction of the trapped atoms towards the photonic waveguide is realised through van der Waals forces alone. According to estimates, this could result in the short-term fractional frequency stability of the strontium atom-chip optical clocks being limited to ~10^{-16} per second, with an accuracy of about 5 x 10^{-15} for fractional intensity stability in the trapping light fields of 0.001.

We consider whether the translation Noether current of a linearised metric spin-2 theory admits a strictly gauge-invariant local representative. For Fierz-Pauli and dynamically regular pure curvature-squared theories in four dimensions, we show that no such representative exists under the stated locality, polynomiality and covariance assumptions. A direct enumeration first excludes nonzero symmetric rank-two tensors at the dimension and derivative order of the quadratic action: in the curvature-squared case, the parity-even conservation kernel consists of the identically vanishing Lanczos–Lovelock combination, while both parity-odd coefficients vanish. Homogeneity of the actions in total derivative number then allows projection by field degree and derivative number, extending the exclusion to all finite-order local polynomial representatives with constant coefficients and no explicit coordinate dependence. The translation current nevertheless defines a nonzero conserved-current class. Independently, a constant translation is a Killing reducibility parameter of the linearised gauge symmetry and determines a nonzero linear surface-charge class, which is distinct from the quadratic energy current. A separate calculation shows why a direct compensator cancellation available in Maxwell theory fails for the metric spin-2 field, although this calculation is not needed for the current-class result. Maxwell theory, propagating free $p$-forms and Yang–Mills theory linearised about the trivial vacuum provide comparisons in which translations are not reducibility parameters and strictly gauge-invariant stress tensors exist. Gauss–Bonnet and pure flat-space $R^2$ gravity lie outside the regular scope of the result, through variational triviality and dynamical degeneracy, respectively.

In this paper, by using signs rule for a class of special power series and for the Laplace integrals, we find the best power mean bounds for the hypergeometric function (sharp at the right endpoint). Our results contain many known conclusions.

The continuous evolution of cloud gaming has enabled users to play high-quality cloud games on thin clients, such as smartphones and tablets. However, how to effectively transmit the gameplay frames with high resolution and low latency, especially over wireless networks, remains a significant challenge. To address this challenge, we propose a Deep Reinforcement Learning (DRL)-based Interest-Aware Adaptive Bitrate (ABR) scheme, called DIA, that guarantees the quality of perceptually key objects in each frame, while dynamically allocating lower resolutions for non-key objects to conserve bandwidth under constrained network conditions. By leveraging an off-the-shelf semantic segmentation model, the proposed DIA scheme can adjust the bitrate at the object-level. We also design a transformer-based feature extractor and an interest-aware reward function to enable the proposed DIA scheme to identify key objects and dynamically adjust their bitrates. This adaptation is guided by users' in-game movement speed, object features, and prevailing network conditions and operates without requiring additional hardware, such as gaze trackers. Extensive experiments show that the proposed DIA scheme outperforms existing ABR schemes in dynamic wireless network environments. Notably, DIA achieves at least a 20% higher normalized average Quality of Experience (QoE) than the State-Of-The-Art (SOTA) ABR algorithms in 5G scenarios.

In test and evaluation (T&E) of AI models, evaluators typically have access only to a held-out test dataset and no knowledge of the training data or its semantic composition. If critical input conditions—such as time-of-day, weather, or scene location—are underrepresented in this test dataset, T&E results can be misleadingly due to this semantic coverage gap. We propose a three-stage generative AI pipeline to detect and fill such semantic gaps without requiring ground-truth labels of these semantics. First, vision-language models (VLMs) perform zero-shot binary labeling of semantic factors, and an embedding-space clustering step identifies which semantics most affect downstream model behavior. Second, diffusion-based image-to-image models transform existing covered images into the underrepresented semantic domain, preserving scene content while augmenting high-level appearance. Third, statistical analysis of the augmented image embeddings verifies that the synthetic imagery improves dataset coverage before it is incorporated. Crucially, the synthetic data is used only to enrich the test set and no model retraining is performed for T&E. We validate the full pipeline on both image classification and object detection under day/night semantic gaps, showing that the augmented test datasets better distinguish models with and without exposure to the missing semantic, thereby yielding T&E conclusions that more closely match those of an ideal, fully covered dataset. Code available at https://github.com/Hashe037/DEDUCE

Earthquake preparation involves progressive damage accumulation, strain localization, and changes in the organization of seismicity. We investigate preparatory processes preceding seven episodes of major earthquakes in the Kuril - Kamchatka subduction zone during 2014 - 2025, including the 29 July 2025 MW 8.8 Kamchatka megathrust earthquake. We track damage evolution using the seismic regularity coefficient dc and seismic strain dynamics coefficient SSD, defined for seismic series. Before 2022, each episode of strong seismicity was preceded by a pronounced dc minimum, commonly accompanied by an SSD minimum, followed by an increase in dc toward a maximum near the strong event or event group. After 2022, dc minima became shallower and were superimposed on a sustained upward trend. We interpret this transition as evidence that the metrics recorded two superimposed processes. The cycles before 2022 reflect complete strain-localization histories on southern or more distant structures that did not contribute to the segment ruptured in 2025. In contrast, the upward dc trend from 2022, with only shallow minima preceding Episodes 6 and 7, marks coupled behavior within the eventual main rupture segment, also influenced by stress transfer from southern source regions.

Any automation within reliability-critical contexts such as medicine or law has to be transparent and verifiable. Despite their advantages, current LLM-based systems do not meet these requirements due to their inherent black-box nature. Retrieval Augmented Generation (RAG) attempts to ground the generative process in the relevant interpretable documents, thereby reducing the risk of AI 'hallucinations' and potentially enabling verification of the initial generation context. However, as the number and size of potential sources increases, naive retrieval approaches tend to perform worse, ignoring critical context elements. In this paper we aim to tackle this problem in the context of German Legal Code. Using structure-driven parsing to create a Knowledge Graph (KG) for the source document navigation and retrieval, we compare several KG-informed chunking and retrieval strategies and evaluate their effect on the accuracy of retrieved citations.

Resource constraints and imperfect quantum operations remain major challenges for quantum machine learning (QML) on noisy intermediate-scale quantum (NISQ) devices, particularly for fault-diagnosis tasks involving imbalanced and heterogeneous data. This work proposes QTN-RQNN, a Quantum Tensor Network-Based Recurrent Quantum Neural Network that integrates tensor-network-inspired nearest-neighbor interactions with recurrent quantum processing. The model sequentially encodes input features while propagating quantum memory across recurrent steps and incorporates QSMOTE for class balancing. QTN-RQNN is evaluated on four heterogeneous fault-diagnosis datasets SPID, CWRUBD, EFDD, and SFDD and compared with QNN, QTN-QNN, and RQNN baselines through classification, ablation, multi-seed, convergence, and quantum-resource analyses. Results show dataset-dependent predictive behavior, indicating that increased quantum architectural complexity does not universally improve classification. On SPID, QTN-RQNN improves the RQNN F1-score from 0.3077 to 0.4516 and ROC-AUC from 0.4485 to 0.6970, while the remaining datasets exhibit different performance-resource trade-offs and sensitivity to recurrent timesteps and circuit depth. Across all datasets, QTN-RQNN reduces the two-qubit-gate count from 32 to 24 per sample relative to RQNN, corresponding to a 25% reduction without increasing the four-qubit requirement; the same relative reduction is obtained for QTN-QNN over QNN. These findings establish QTN-RQNN as a resource-conscious recurrent quantum architecture and highlight the need to jointly consider predictive performance and circuit-resource requirements for near-term quantum fault diagnosis.

Neural networks (NNs), while highly flexible and versatile, are often poorly calibrated. Post-hoc calibration methods adjust the outputs of an already trained model using a separate calibration dataset and additional calibration parameters. This process adds overhead that requires separate optimization of the calibrator and subsequent model selection. In this paper, we show that standard (post-hoc) calibrators can be trained end-to-end with the backbone via a bilevel Stackelberg formulation. This formulation eliminates the need for a separate calibration set or a dedicated calibration loss by jointly optimizing accuracy and calibration using cross-entropy, introducing only one additional hyperparameter: the calibrator learning rate. Although our framework supports calibrators ranging from temperature and matrix scaling to NN layers, our results show that using the final classification layer itself as the calibrator yields accurate, well-calibrated models with no additional parameters. Our approach outperforms SOTA calibrators with minimal added complexity, while also avoiding the multiplicative overhead of separate calibration components in settings such as multi-task learning.

Standard Policy Gradient (PG) algorithms suffer from high estimation variance because they rely on Monte Carlo updates. While off-policy variance reduction mitigates this problem, establishing optimal sample complexity currently requires either restrictive assumptions, such as uniformly bounded importance weights, or impractical second-order operations. In this work, we introduce Loopless Variance-Reduced PG (LVR-PG), a unified theoretical framework that abstracts a class of first-order estimators that integrate recursive updates with defensive importance sampling, including known estimators such as PAGE-PG and STORM-PG. We prove that this algorithmic class achieves $\tilde{\mathcal{O}}(\epsilon^{-2})$ last-iterate global convergence under weak gradient domination for general parametric policies. We establish this guarantee using purely first-order updates, with minimal assumptions and without gradient preconditioners. By extending state-of-the-art lower bounds to continuous parametric policies, we establish an $\Omega(\epsilon^{-2})$ sample complexity limit, confirming our framework is minimax-optimal up to logarithmic factors and horizon terms.

Let $r_1,\ldots,r_t$ be positive integers, and let $\mathcal A$ be the set of integers of the form $2^{k_1^{r_1}}+\cdots+2^{k_t^{r_t}}$, where $k_1,\ldots,k_t$ are positive integers. For any positive integer $n$, let $r(n)$ be the number of representations $n=p+a$ with $p$ an odd prime and $a\in \mathcal A$. We prove that if $r_1^{-1}+\cdots+r_t^{-1}=1$, then, for every fixed positive integer $m$, the set of integers $n$ with $r(n)\ge m$ has positive lower asymptotic density. In particular, this confirms a conjecture of Chen.

Diffusion models typically learn the generative dynamics of a single target distribution or a family of distributions indexed by conditioning variables fixed during training. This limits their ability to adapt to a previously unseen distribution from only a small set of examples. We introduce an in-context solution operator learning framework for diffusion models, in which each target distribution is treated as a task drawn from a meta-distribution and is specified at inference time by an empirical context set. Our method learns an in-context score operator that maps the context samples, a noisy query point, and the diffusion time to the task-dependent score, enabling reverse-diffusion sampling on unseen tasks without retraining or parameter updates. A key observation is that the task-dependent score admits a Gaussian-kernel conditional-mean representation. Its empirical Nadaraya–Watson approximation can be represented exactly by a model in our constructed transformer class, providing a principled foundation for the proposed query-dependent attention architecture. The operator is trained through amortized denoising score matching across task distributions. We establish generalization-error bounds that quantify the effects of the context size, the numbers of training tasks and within-task samples, model complexity, and intrinsic and ambient dimensions. Numerically, we evaluate the method on a range of scientific and visual datasets, including Navier-Stokes, ERA5 climate data, QuickDraw sketches, and Food-101. The results show that the proposed method performs well on relatively simple task families, while its performance degrades as task complexity increases, thereby illustrating both the promise and the limitations of in-context learning for diffusion models.

Learning how local perturbations affect a network is important for identifying which components drive or absorb a system's response. We formulate this as seed-conditioned response ranking: given a perturbation at one seed node, ranking all other nodes by response strength. This is challenging when the model is given only the interaction topology, while the mechanism governing how perturbations propagate must be learned from data. We present VQIL (Visible Quantum Interference Learning), a trainable Szegedy quantum walk for seed-conditioned ranking. Edge weights, link phases, and per-node phase rotations are optimized end-to-end by backpropagation while the walk operator remains unitary throughout training. Node scores are obtained from a geometrically discounted coherent sum of quantum-walk amplitudes, enabling quantum interference before the final squared-norm readout. Because each learned parameter is attached to a named edge or node, the trained model provides directly inspectable hypotheses about the roles of individual graph components in the propagation mechanism. On AC power-flow perturbations over six transmission grids, with electrical parameters withheld from the model, VQIL improves Spearman rank correlation from 0.147 for the untrained walk to 0.831 and outperforms an identically trained random-walk model. On gene perturbation over five protein-protein interaction networks with Perturb-seq supervision, VQIL improves rank correlation from 0.171 to 0.620, outperforming the strongest learned classical and neural baselines. These results show that trainable quantum-walk interference provides a scalable and interpretable inductive bias for learning seed-conditioned response rankings on graphs.

Harvesting technologies have evolved from conventional mechanized systems based on bulk mechanical interaction toward increasingly automated and robotic systems capable of perception, decision-making, and localized actuation. This evolution has also created a classification problem, as harvesting systems are commonly described according to crop type, mechanical removal principle, sensing technology, robotic architecture, or degree of automation, making direct comparison difficult. This paper proposes a unified taxonomy for agricultural harvesting technologies based on two primary classification axes: the physical principle governing plant-machine interaction and the level of automation incorporated into the harvesting process. Selectivity is treated as an emergent functional attribute that expresses the biological and spatial scale at which perception and decision-making influence physical actuation. Conventional mechanized methods, including cutting, pulling, vibration, impact, brushing, pneumatic action, and combined systems, are analyzed together with semi-automated and robotic harvesting approaches. The analysis shows that increasing automation does not necessarily result in greater selectivity. A qualitative transition occurs when perception and decision-making directly influence product-level actuation, shifting harvesting from distributed or bulk interaction toward localized and selective manipulation. The proposed framework provides a common conceptual basis for positioning and comparing mechanized, automated, and robotic harvesting technologies. By separating the physical mechanism of product removal from the automation level and resulting selectivity, the taxonomy clarifies technological boundaries, identifies intermediate configurations, and supports the analysis and development of future harvesting systems.

Audio-visual dataset distillation aims to synthesize a small dataset from a large paired audio-visual dataset, while nearly preserving the training utility for neural networks. Existing approaches such as DM and AVDD face two critical limitations: 1) they are memory-intensive, making them difficult to scale to complex architectures and higher samples-per-class (SPC) settings, and 2) they often fail to preserve strong cross-modal semantic correspondence. To tackle these issues, we propose X-AVDD, a cross-attention-based distillation method that explicitly captures inter-modal relationships during synthesis. By decoupling cross-modal alignment from joint optimization, X-AVDD is substantially more memory-efficient, reducing end-to-end synthesis cost by $\sim 8\times$ GPU-hours, while producing highly aligned synthetic data. Empirically, X-AVDD significantly outperforms other audio-visual distillation methods. On VGGSound, an audio-visual dataset of short YouTube clips, X-AVDD improves top-1 test accuracy from $8.2\%\rightarrow24.0\%$ at SPC=$10$ and $9.8\%\rightarrow26.5\%$ at SPC=$20$. Furthermore, X-AVDD improves cross-architecture generalization (e.g., $25.5\%\rightarrow35.3\%$ for a ViT target on AVE dataset at SPC $=50$) and successfully scales to complex architectures such as AudioCLIP, achieving $62.3\%$ accuracy using only $5.65\%$ of the full dataset, against the full dataset accuracy of $72.4\%$ while completing in $9\times$ lower training time compared to the full dataset.

The PLATO mission, planned to launch in spring 2027, will observe a fixed Southern field, LOPS2, for at least 2 years with the goal of discovering and deriving precise parameters for planets as small as Earth around FGK dwarfs and sub-giants. Afterwards, PLATO could continue to monitor the same field or move to a Northern field, LOPN1, overlapping with the Kepler field. We assess these two observing strategies in terms of sensitivity of detecting small and long period planets, and subsequently we estimate yields for both observing strategies. Within this we account for including the Kepler data in the LOPN1 field, monitoring of up to 8 years of the LOPS2 field, and the sensitivity improvement from combining TESS and PLATO data since both fields have been and will be continued to be observed with TESS. We find that by monitoring both fields, and hence more stars, the total number of newly discovered planets is maximised. However, to optimise the number of discovered small and long periods planets, long-term monitoring of one field produces the best results. The combination of Kepler and PLATO data improves detection sensitivities; however, as this only covers a fraction of the LOPN1 field and stellar sample, it improves the yields of small and long period planets only marginally. Hence, to detect and characterise small and long period planets in this sparsely explored exoplanet parameter space, a longer continuous observation of the LOPS2 field is best.

We study the design of a statistical privacy mechanism under a bounded $\alpha$-lift leakage constraint. An agent observes useful data $Y$, which is arbitrarily correlated with private data $X$, and releases $U$. The goal is to maximize Pearson $\chi^2$-information $\chi^2(Y;U)$ while satisfying the point-wise privacy constraint. This utility is an $f$-information that is nonnegative, obeys data processing, and measures the aggregate normalized change from the prior $P_Y$ to the posteriors $P_{Y|U=u}$. The main benefit of this choice is that $\chi^2(Y;U)$ is exactly quadratic in the posterior perturbations; hence, no local approximation of the utility is required. For the nonlinear $\alpha$-lift constraint, we derive a uniform Taylor-remainder bound that remains valid for the optimizing, $\epsilon$-dependent perturbation. When $P_{X|Y}$ is invertible, this yields a closed-form quadratic surrogate, an explicitly scaled binary mechanism that satisfies the original $\alpha$-lift constraint, a computable sufficient range for $\epsilon$, and an $O(\epsilon^{3/2})$ bound on the utility gap. For $\alpha=2$, the solution is exact over an explicit range. We transfer these results to point-wise Rényi-divergence privacy, including an exact-feasible mechanism, an explicit privacy-budget range, and matching gap bounds. We also give an exact formulation for arbitrary leakage matrices and evaluate exact-feasible mechanisms numerically.

Simulations of the evolution of matter on static metric backgrounds allow us to estimate backreaction effects like dynamical friction at first order in the matter density. Previous simulations of dynamical friction have used boosted isotropic and quasi-isotropic Kerr coordinates, which simplifies the extraction of the relevant diagnostics but results in a more complicated implementation of the evolution. In particular, the lack of horizon penetrating coordinates means that high resolutions are needed to ensure convergence, and the boosted, spinning black hole metric cannot easily be written in a closed form, making it more costly to evaluate. In this paper we describe how to reformulate the calculation of the diagnostics in the boosted Kerr-Schild frame for the non-spinning case. We verify that our implementation reproduces previous results quantifying dynamical friction in wave-like dark matter in the appropriate regimes. We discuss the subtleties of transforming diagnostic observables between frames in a 3+1 evolution and develop a general framework for this process.

Accurate monitoring of power flow is fundamental to the secure and reliable operation of modern power systems. Existing data-driven power flow estimation methods typically assume that synchronized measurements are available at every bus, requiring extensive deployment of phasor measurement units (PMUs). Such dense sensing infrastructures substantially increase installation, communication, and maintenance costs. This paper proposes a learning-based framework that minimizes the number of installed PMUs while maintaining high-accuracy systemwide power flow estimation. Instead of relying on measurements from all buses, the proposed method identifies an informative subset of buses whose synchronized measurements are sufficient to accurately reconstruct branch power flow across the network. To this end, a neural network is developed to learn the nonlinear relationship between PMU measurements and system-wide power flows, and multiple PMU deployment scenarios are evaluated to quantify the trade-off between measurement availability and estimation accuracy. The results enable utilities to identify cost-effective PMU placement strategies that balance deployment cost and prediction performance. Case studies on the IEEE 30-bus test system demonstrate that the proposed framework significantly reduces the number of required PMUs while maintaining accurate power flow estimation.