Papers

We modify the approach to the arithmetical form of the large sieve by relying on the Parseval identity rather than on an approximate Bessel inequality and as a consequence, we prove that sifted sets are either small or the attached Fourier polynomial is very regular at the origin. We also discuss the optimality of this approach and shall show in passing that $N\int_{-\delta}^\delta|S(\beta)|^2d\beta\ge(1-2/\sqrt{\delta N})|S(0)|^2$ for any trigonometric polynomial of length~$N$.

Current source inverters (CSIs) are emerging as competitive alternatives to voltage source inverters (VSIs) for high-speed motor drives because they inherently suppress harmonic currents, offer output-voltage boosting, and reduce motor-side electromagnetic interference. Sensorless control of CSI-fed permanent magnet synchronous machines (PMSMs) remains less mature than VSI counterparts, primarily because conventional implementations require measurement of all circuit state variables. This paper proposes a weak-resonance-based approximation that reconstructs stator currents from inverter modulation commands and feedforward capacitor-current estimation, thereby reducing the sensing requirement to the dc-link current and two terminal voltages. The proposed observer–PLL structure is analysed for computational complexity and validated through simulations and experiments on a 100~W, 100~krpm GaN-based two-stage CSI prototype (buck + CSI). Quantitative results demonstrate accurate rotor-position tracking and high-bandwidth operation while maintaining the reduced sensor set. Proposed method performs well in simulation and experiment, achieving a position estimation error below $5^\circ$ at 100~krpm.

We investigate Anderson localization in a one-dimensional lattice with non-Hermitian off-diagonal quasiperiodic disorder, extending a recently studied Hermitian mosaic model to the non-Hermitian regime. Using Avila's global theory, we derive the exact Lyapunov exponent and the complete phase diagram in the complex energy plane. This work contains two central findings. First, we discover mobility arcs—open curved segments in the complex plane—as a new class of mobility edges and the generic form of open mobility edges, which coexist with closed mobility rings in a complementary parameter regime. These arcs share the same localization physics as the previously reported mobility lines: eigenstates are delocalized if and only if their energies lie exactly on these sets; any deviation yields localized states. This constitutes a striking measure-zero delocalization phenomenon: delocalized states occupy only zero-measure sets (arcs or lines) in the complex plane, in sharp contrast to the mobility rings, which enclose a finite-area region of delocalized states. Second, we reveal that mobility rings, arcs, and lines all share a common mathematical origin in the generalized Joukowski transformation $P(E) = \frac{1}{2}(u - w^2/u)$, rooted in the algebraic structure of the underlying polynomial: the preimage of the boundary of an elliptical region under the polynomial map $P(E)$ gives the rings, while the branch cut inside this ellipse gives rise to the mobility arcs and lines in the complementary parameter regime.

The connection between the Kolmogorov-Arnold representation theorem (KART) and neural network design has led to the development of Kolmogorov-Arnold Networks (KANs), with applications ranging from STEM problems to AI tasks. In this paper, we investigate the effectiveness of a KAN variant, SechKAN, which relies on hyperbolic secant (sech) functions as basis functions, with a 1D projection to reduce the number of parameters to a level comparable to MLPs. We evaluate SechKAN on three 1D classification datasets: UCI Human Activity Recognition (UCI HAR), ElectricDevices, and Crop, and compare it with several effective networks, including EfficientKAN, MLP, CNN1D, ResNet1D, and DSCNN1D, using approximately comparable parameter budgets. The results indicate that SechKAN achieves competitive performance across the three datasets, with particularly strong performance on Crop. Ablation studies further show that grid size and normalization affect performance, suggesting that SechKAN's effectiveness depends on the dataset and architectural choices. Our source code and experimental implementation are publicly available at: https://github.com/hoangthangta/SechKAN_1D.

The static Love numbers of four-dimensional Schwarzschild black holes vanish classically. We show that this is not true in the quantum theory. Working in the worldline effective field theory, we compute the leading quantum correction to the Compton amplitude for massless scalars and photons scattering off a Schwarzschild black hole at first post-Minkowskian order, and find logarithmic divergences associated with the renormalization of the static Love numbers. We conclude that renormalization-group running necessarily generates Love numbers suppressed by the square of the Planck length in units of the horizon radius, $\ell_{\rm pl}^{2}/r_{\rm h}^{2}$. This is consistent with the classical vanishing being a consequence of an accidental symmetry of static general relativity, broken by quantum loops. Extrapolated to gravitational perturbations, our results suggest that the quadrupolar static Love numbers of Schwarzschild are of size $\lambda \sim M r_{\rm h}^{2} \ell_{\rm pl}^{2}$.

Deep neural networks approximate functions by composing affine maps with nonlinear activations, but how composition itself creates approximation power is not yet fully understood. We investigate a fundamental mechanism: geometrically weighted sums of iterates of a single scalar generator function. This mechanism underpins the classical tent-map construction of the function \(x - x^2\) and related recursive representations used by Yarotsky, W. E, et al., to analyze the approximation powers of deep neural networks. First, we establish a rigidity theorem: for continuous piecewise linear generators with a finite number of segments, any \(C^3\) function that can be represented in this way is at most quadratic. For non-affine quadratic functions, the geometric factor is at least $1/4$. This result both reveals limitations of the tent-map approach and complements existing methods based on hierarchical bases and recursive polynomial constructions. Second, using an exact remainder identity as guidance, we construct a smooth generator whose iterates yield doubly exponential error decay in total depth for square approximation and, through multiplication modules, for each fixed polynomial. For power series with absolutely summable coefficients on \([-1,1]^d\), distributing depth according to monomial degree yields a uniform approximation error of order \(O(e^{-cL^{1/d}})\) on each interior cube. These findings demonstrate how generator dynamics and remainder estimates govern depth allocation and approximation rates of deep neural networks.

Analog circuit sizing remains a challenging and time-consuming task due to the large design space, strong performance trade-offs, and increasing circuit complexity in scaled technologies. Although recent large language model (LLM)-based methods show promise in improving sample efficiency and interpretability, existing approaches often lack explicit circuit-topology understanding and are mainly evaluated on relatively simple analog building blocks. This paper presents a multi-agent LLM-based framework for complex analog circuit sizing. The proposed framework first analyzes the circuit topology and decomposes the netlist into functional blocks and substructures. It also extracts lightweight design knowledge for reuse. Based on the extracted topology and knowledge, a planner coordinates multiple role-specialized sizing agents to update design variables and achieve global performance specifications. This workflow mimics the collaborative process of an expert analog design team and provides a structured, interpretable, and simulation-driven optimization procedure. The framework was validated on eight circuits, with the largest design containing up to 55 transistors and 60 sizing variables. Notably, for the LDO benchmark, the proposed method achieved a 60% success rate with an average of 83 iterations, where classical optimizers failed to find feasible solutions. Further, ablation studies demonstrate that topology understanding, design-knowledge infusion, and agent specialization provide complementary benefits. The source code is available to support reproducibility.

The occurrence of chaos in multipendular FP resonators (without light propagation delay) is illustrated. The probability of evolving a chaotic solution in a given time, starting from initial conditions contained in a neighbourhood of a stable equilibrium point, is estimated, and its dependence on the most relevant system parameters is investigated. The possible relevance of these results for interferometric gravitational wave antennas is discussed.

Using operator-valued subordination, we derive a formula for the $2$-norm distance on reflexive lattices generated by double triangles of projections under freeness assumptions on their graph coefficients, and provide explicit evaluations in both the scalar and free-projection cases.

The proposed high-luminosity Super Tau-Charm Facility (STCF) offers a clean experimental environment for threshold studies of exotic hadrons. In this Letter, we evaluate the expected significance for the fully-charmed vector tetraquark candidate $T_{4c}$ in the $e^+e^- \to \eta_c J/\psi$ channel at the STCF. Using Monte Carlo simulations of an energy scan from $\sqrt{s}=6.71$ to $6.79~\mathrm{GeV}$ with $100~\mathrm{fb}^{-1}$ per point, we adopt a high-efficiency single-tag (ST) reconstruction of $J/\psi$ leptonic decays as the primary strategy, with a double-tag (DT) reconstruction of $\eta_c \to K^+K^-\pi^0$ retained as an independent cross-check. Because the signal cross section is governed by the still-uncertain dielectron width $\Gamma_{ee}$, we consider three benchmark hypotheses, $\Gamma_{ee}=0.25, 0.5, 1~\mathrm{eV}$. The corresponding expected ST significances are $5.1\,\sigma$, $10.6\,\sigma$, and $20.5\,\sigma$, respectively. These results indicate that the STCF can provide meaningful sensitivity to fully-charmed tetraquark states near threshold.

Attention computation makes inference expensive in video diffusion transformers (vDiTs), which generate videos through iterative denoising. Block-sparse attention (BSA) reduces this cost by computing only blocks selected by a logical mask, which specifies attention interactions to compute. However, coupling logical block geometry to execution choices limits adaptation to varying masks and graphics processing units (GPUs), while runtime kernel specialization can incur preparation overhead that outweighs execution time savings. We present Tessera, a specialized runtime for dynamic BSA that decouples logical masks from GPU execution while preserving specified attention interactions. Its physical mapping layer retains, combines, or subdivides logical attention blocks into physical tiles suited to different attention mask shapes and GPU architectures. Its task organization layer groups and schedules tiles within GPU tasks to reuse data, expose parallelism, and overlap data movement with computation. Finally, profile-guided regime selection enables low- overhead execution plan selection through a lookup table constructed from offline profiling. We implement Tessera with specialized CUDA kernels supporting four NVIDIA GPU generations. Evaluated on 2,315 real attention masks and industrial video diffusion models, Tessera achieves up to 6.79x BSA request speedup over baseline systems in the evaluated video diffusion models.

In this paper we extend the description of spatial topological chaos for bistable lattice equations in a small diffusion regime. We generalize the Keener's proof and show that stationary solutions are topologically conjugate to biinfinite sequences on a three-symbol alphabet. We provide a priori estimates for the values of stationary solutions and use them to show that there are exactly two monotone stationary fronts (up to a translation). Finally, we apply the localization of all stationary solutions to characterize their stability and get the existence of unique stable and unstable monotone fronts (up to a translation).

We investigate how the chiral anomaly influences the decomposition of nucleon spin between quark spin and orbital motion. We extend the chiral quark soliton model by including a flavor-singlet pseudoscalar field whose mass is generated by the chiral anomaly. Within this framework, we calculate the quark and antiquark helicity and orbital angular momentum distributions, accounting for the singlet field induced by the rotation of the soliton. We vary the anomaly-induced mass while keeping the remaining model parameters fixed to examine how the anomaly affects the spin decomposition. The induced field enhances the quark helicity contribution relative to the conventional pion-only model. Increasing the mass suppresses this enhancement and increases the orbital contribution, while the two contributions continue to account for the nucleon spin. This redistribution remains modest over the parameter values considered and does not qualitatively alter the spin decomposition of the conventional model.

We develop Hachimori's study on the unramified Iwasawa modules using extensions with restricted $p$-ramification. Let $K$ be an algebraic number field, and $p$ a prime number which splits into two distinct primes $\mathfrak{p}$, $\mathfrak{p}'$ in $K$. Assume that $K$ and $p$ satisfies several (somewhat strict) conditions. Let $K_\infty /K$ be the cyclotomic $\mathbb{Z}_p$-extension. In the present paper, we give a method to study the structure of the unramified Iwasawa module $X (K_\infty)$ by using abelian $p$-extensions unramified outside $\mathfrak{p}$. We give a sufficient condition for $|X (K_\infty)|$ to be finite in terms of such extensions. We also give a similar criterion for $X (K_\infty)$ to be finitely generated over $\mathbb{Z}_p$. In the latter part of the present paper, we consider the case where $k = \mathbb{Q} (\sqrt{q})$ with an odd prime number $q$ and apply our results to the cyclotomic $\mathbb{Z}_2$-extension $k_\infty /k$. We give a necessary and sufficient condition for $X (k_\infty)$ to be cyclic over $\mathbb{Z}_2$, which is different from the former results given by either Mouhib-Movahhedi or Mizusawa-Mouhib. We also give several sufficient conditions for the validity of Greenberg's conjecture.

State production and its subsequent decay can each require a different angular-momentum coupling treatment. Although long recognized, this distinction is often overlooked in the two-step model, where $LS$ coupling is traditionally applied to both production and decay ($LS/LS$), even when the decay requires $LSJ$ coupling. We introduce a hybrid $LS/LSJ$ formulation that uses established angular-momentum recoupling theory to project the initial $LS$ alignment onto the fine-structure $J$ states, evaluates their decay in $LSJ$ coupling, and, for well-separated $J$ levels, sums their contributions incoherently. The $1s2s2p\,{}^{4}\!P_J$ manifold produced by single-electron capture in $\mathrm{C}^{4+}(1s2s\,{}^{3}\!S)+\mathrm{He}$ collisions provides a good test case, with $M_L$-resolved theoretical production cross sections and absolute measurements of the Auger single-differential cross section available. Using the same $LS$ production cross sections in both treatments, we find that the hybrid $LS/LSJ$ treatment strongly suppresses and can even invert the Auger anisotropy relative to the traditional $LS/LS$ treatment, substantially improving agreement with the measurements. Such an incoherent hybrid formulation provides the appropriate framework when $LS$ coupling governs the initial interaction dynamics while well-separated fine-structure levels require an $LSJ$ description of the decay.

Visual object removal can eliminate a target from video frames, yet its acoustic trace persists in the soundtrack, causing obvious audio-visual inconsistency. Existing video inpainting models operate solely on pixels, while audio editing models, especially for the sound removal task, are typically driven by text and therefore rely on limited single-modal control, which is less effective than multimodal guidance that provides stronger semantic grounding and temporal synchronization cues. In this paper, we present Text-Visual Guided Sound Removal (TV-AudioRemover), a target sound removal framework that leverages the visually edited video together with a natural-language instruction to suppress the sound associated with the removed visual object from the original audio mixture. To acquire high-quality training data, we devise a pipeline to construct a million-scale dataset of single-object audio-visual aligned samples, from which we synthesize mixture-target pairs customized for model training. To effectively leverage visual context and follow instruction intent, we augment the model architecture with task tokens, generalizable instruction modeling, and modality-specific global guidance. We further adopt multi-task training to strengthen task-role comprehension, and employ a hard-mixture curriculum that leverages semantically similar acoustic mixtures during fine-tuning to enhance fine-grained source discrimination. To support evaluation, we present AV-Remove-Bench, a comprehensive audio-visual object removal benchmark, along with dedicated objective metrics and an MLLM-based evaluation protocol. Experiments demonstrate that our method achieves state-of-the-art performance on both subjective and objective metrics. Project page: https://yjx-research.github.io/TV-AudioRemover/.

The Ehrhart series of the $n$th Birkhoff polytope is $\sum_{r\geq0}H_n(r)z^r$, where $H_n(r)$ counts nonnegative integer $n\times n$ matrices whose row and column sums all equal $r$. We present an exact method for computing this series using constant terms and finite fields. A root filter expresses $H_n(r)$ as a weighted sum of the values $h_r(M)^n$, where $h_r$ is the complete homogeneous symmetric polynomial and $M$ ranges over multisets of $m$th roots of unity with $m=r+1$. Constant-term cancellation reduces the evaluation of $h_r(M)$ to a sum over repeated elements $a$ of $M$. For a particular $a$ of multiplicity $\mu_a$, the computation uses a generalized Todd coefficient of degree $\mu_a-2$. Sums of $h_r(M)^n$ over selected multiplicity classes are handled using symmetric function techniques. Together with the remaining individual evaluations, this gives $O_n(m^{n-5}+m^4)$ field operations for each admissible prime and fixed $n\geq5$. An explicit bound and the Chinese remainder theorem recover the integer counts, and Ehrhart symmetry determines the full series. The same method applies to the World Cup problem, which counts the same matrices with diagonal entries required to be $0$. We prove correctness and compute complete series for both families through order $13$. The Birkhoff series for orders $10$–$13$ and the World Cup series for orders $9$–$13$ are tabulated in the appendices.

We present the first experiments on isolated sign language recognition (ISLR) for Icelandic Sign Language (ÍTM). We use ÍTM SignWiki, a dataset derived from a bilingual Icelandic–ÍTM online dictionary. It is genuinely low-resource: 1,845 videos cover 849 classes, 86% of which have only two examples, making the full task effectively one-shot recognition across signers. We compare two open-source ISLR frameworks, OpenHands and SPOTER, on three tasks of increasing vocabulary size (22, 117 and 849 classes), and evaluate three pose estimators and two forms of cross-lingual transfer. With ÍTM data alone, SPOTER outperforms OpenHands on all three tasks, and MediaPipe poses give better results than AlphaPose or SDPose. Cross-lingual transfer brings the largest gains: pretraining SPOTER on American Sign Language data before finetuning on ÍTM raises accuracy by 14–24 percentage points, to 72.7%, 47.9% and 22.6% on the three tasks, and multilingual training with data from six other sign languages lifts OpenHands from 1.41% to 28.86% on the full task. Although far from practical use, the results suggest that transfer from better-resourced sign languages is promising for very low-resource ones. We release our adapted versions of both frameworks.

Imperfect photon sources degrade the performance of interference-based linear-optical entangling gates. We derive a universal fidelity law for these gates in terms of two standard source metrics: Hong-Ou-Mandel visibility and second-order coherence. We model the multiphoton component as an arbitrary overlap between noise and signal photons. While different values of the signal-noise overlap produce different error structures, associated with different physical platforms, all result in the same fidelity law. We test the model on our actively synchronized CNOT-gate experiment and against published results across different platforms, with no free parameters. This framework turns source characterization into a unified benchmark for photonic quantum computing.

Sparse instance representations provide a compact interface for spatial LiDAR-camera and temporal past-current interaction in multimodal perception and E2EAD. Effective interaction requires reliable instance correspondences despite geometric discrepancies and heterogeneous semantic representations. Attention-based methods exploit contextual semantics but often require specialized representation alignment, increasing computational overhead. In contrast, association based on structured object states is efficient and interpretable but lacks contextual evidence to resolve ambiguous matches. To combine these complementary strengths, we propose MatchFusion, a learnable instance matching and fusion module for spatio-temporal multimodal autonomous driving. MatchFusion initializes pairwise affinities using geometric similarity and category consistency, then selectively refines structurally plausible associations using instance embeddings. The resulting soft matchmap guides a common residual aggregation operator for adaptive information exchange. This unified matching-fusion formulation supports spatial LiDAR-camera and temporal past-current interaction, using multi-view image-plane geometry and motion-compensated BEV geometry as the respective structural priors. Experiments on nuScenes demonstrate consistent perception gains across diverse front-end configurations. Compared with a prior instance-centric fusion method, the MatchFusion-equipped system achieves higher perception accuracy while reducing FLOPs by 55.3% and GPU memory usage by 39.3%, with the matching-fusion module accounting for only 3.7% of total perception latency. Integrating temporal MatchFusion into SparseDrive further improves perception within an E2E framework without additional supervision. These results establish explicit-implicit matching as an effective and efficient mechanism for spatio-temporal instance interaction.

Biomedical Entity Linking disambiguates mentions to entities in a knowledge base (KB), making it the cornerstone of information extraction pipelines. While embedding-based models are a popular approach for the task, they suffer from a key limitation. They compress mentions (and entities) into a single vector, forcing the model to average away crucial fine-grained differences. We present BELXTR, a novel embedding model based on the multi-vector (a.k.a. late interaction) architecture, which allows to leverage token-level matching information. BELXTR extends the original XTR model to biomedical entity linking by integrating an existing task-specific training objective and exploring active query expansion. Experiments across ten corpora and five KBs show that BELXTR improves upon current state-of-the-art in half of the corpora with an average improvement of 5pp recall@1. The largest gains are reported on the challenging cross-species gene disambiguation subtask, where BELXTR outperforms an LLM-powered retrieve-and-rerank pipeline and closely approaches a specialized rule-based system. Our results highlight multi-vector models as a practical alternative to hard-to-maintain rule-based systems or in scenarios where LLM-based reranking is too costly as in PubMed-scale mining. The code to reproduce our experiments can be found at: https://github.com/sg-wbi/belxtr.

In this paper, we investigate the elastic inverse scattering problem in a two-dimensional cavity with an impedance boundary condition. We prove that the scattered fields inside the cavity satisfy a reciprocity relation, and prove that both the shape of cavity and the impedance parameter on the boundary can be uniquely determined from the scattered fields measured on an interior curve. Numerically, we first employ the linear sampling method to reconstruct the shape, and then recover the impedance parameter by combining potential theory with the least squares method. Several numerical experiments are presented to demonstrate the feasibility and effectiveness of our methods.

We numerically construct the Sasaki–Einstein metric on the link of the cone over the second del Pezzo surface $\mathrm{dP}_2$ and two primitive harmonic basic $(1,1)$-forms at its irregular volume-minimizing Reeb vector. Being toric, the metric in symplectic coordinates is encoded in a single convex function on a polygon. We approximate the correction to the canonical Guillemin potential in two ways: polynomial expansion and neural networks. The polynomial fit achieves a held-out mean-squared Monge–Ampère residual below $10^{-13}$, in contrast to the $10^{-2}$ plateau for the non-volume-minimizing regular Reeb vector. We validate our method against closed-form metrics of $Y^{p,q}$ using curvature invariants, and also against the numerical result of Doran et al. (2007) for the Kähler–Einstein metric of $\mathrm{dP}_3$ using the Laplacian spectrum of low torus-invariant modes. The integrated Riemann-squared curvature of the $\mathrm{dP}_2$ link agrees to two parts in $10^{9}$ with the closed form conjectured from the dual field theory. Our data for the metric and harmonic forms can be used to study warped non-conformal holographic IIB backgrounds, the analogues of the Klebanov–Tseytlin solution on the conifold.

Accurate anomaly detection is essential for reliable and secure operations of microservice systems. While an increasing number of studies have shifted from unimodal modeling to multimodal interaction and fusion, effectively leveraging reliable cross-modal information remains challenging. The challenge primarily stems from two aspects. Firstly, different modalities are influenced by factors like load fluctuations, leading to dynamically changing reliability. Secondly, multimodal data exhibit heterogeneity in both structure and semantics. Therefore, we propose a confidence-guided Cross-Modal knowledge Transfer method for multimodal Anomaly Detection (CMT-AD). It jointly models metrics and logs within a unified deep clustering framework and estimates modality reliability through the soft clustering distributions, where clustering uncertainty is quantified into confidence scores. Guided by these confidence scores, the model actively analyzes the contributions of each modality in cross-modal interactions and supplements low-confidence modalities with knowledge from high-confidence ones. To further mitigate cross-modal heterogeneity, we introduce a gated intermediate modality and design structural and semantic consistency constraints that align the original modalities with the intermediate modality to preserve similarity structures and semantic distributions across modalities. Furthermore, intra-modal and cross-modal regularization terms are incorporated to enhance cluster compactness and mitigate negative transfer. We evaluated CMT-AD on three large-scale datasets, and the results demonstrate that CMT-AD outperforms state-of-the-art approaches and achieves an F1-score higher than 0.9.

We study the strong-shear limit of bistable traveling fronts in infinite cylinders with periodic transverse boundary conditions. We prove that, for every sufficiently regular periodic shear profile, the front speed normalized by the flow amplitude converges as the amplitude tends to infinity. After suitable longitudinal rescaling and translations, the corresponding profiles converge along subsequences to full fronts of the limiting degenerate equation. Furthermore, under a Hörmander-type non-degeneracy condition on the shear, the limiting front is proved to be regular and unique up to translation, and the whole normalized family converges uniformly. A main difficulty is that the sign-changing bistable reaction allows the limiting transition to split through intermediate transverse equilibria. We rule out this possibility by exploiting the instability of such equilibria together with suitable regularization and comparison arguments. Finally, in contrast with the combustion case, where every nonconstant mean-zero shear yields a positive limiting speed, we construct an example showing that, for a fixed bistable reaction, smooth mean-zero shears can produce negative, zero, or positive limiting speeds. In particular, the example shows that a shear which accelerates propagation at small amplitudes may reverse the propagation direction when its amplitude becomes large.

The multirate partial differential equation (PDE) technique is considered as one of the most powerful methods for the simulation of radio frequency (RF) circuits. The method reformulates the underlying ordinary differential algebraic equation (DAE) by a PDE. The DAE solution is obtained along a characteristic curve of the PDE. Numerical solution techniques comprise (envelope modulated) Harmonic Balance (HB), multistep finite difference (FD) and spline/wavelet techniques. All techniques have disadvantages when mixed lumped/distributed devices occur and the waveforms exhibit pulse shaped behavior. Furthermore, communication systems are modeled and simulated by the Equivalent Complex Baseband (ECB) approach. Signals and impulse responses are described by their baseband counterparts, employing the Hilbert transformation. A coupled system/circuit simulation requires therefore novel numerical techniques for the solution of integro-differential equations.

Learned-memory methods store information in an explicit table and consume it through a separate reader, allowing addressing, storage, and reading to be modified independently. We study whether useful memory can also be generated rather than only retrieved. MemoryAthena uses three pathways: direct Engram retrieval (E), generation from retrieved Engram cues (GE), and generation from causal backbone states without consulting the memory table (GH). Generated memory is conditionally useful: it can complement E in one context but interfere with it in another. MemoryAthena therefore treats E as an anchor and learns when a generated representation should intervene. With the backbone, memory, generators, and readers frozen, a lightweight causal routing head is trained from counterfactual future-token likelihood advantages of GE and GH relative to E. At inference time, an admitted candidate modifies the E residual through bounded interpolation, while rejection recovers the direct pathway exactly. On question answering, MemoryAthena raises the five-task average from 37.65 to 39.28 over the direct pathway of the same checkpoint, while the six-task general-NLP average increases from 76.73 to 79.13. The complete memory-side system contains approximately 201M parameters, excluding the frozen backbone. Further analyses show complementary strengths among E, GE, and GH across tasks and inputs. These results support generated memory as a selective correction to direct retrieval and highlight routing when, which, and how strongly to intervene as the central challenge.

Clinical AI can predict outcomes using information generated by existing clinical judgment. Such performance alone cannot establish usefulness before recognition or across care processes. We propose evidence requirements matched to intended use and information conditions, with additional state-reference and proxy-relative timing tests for stronger recognition claims. This framework helps hospitals assess whether the conditions supporting performance hold when assistance is most needed. These questions remain outside current evaluation practice leakage checks, overall predictive performance, and external validation do not address them; the framework makes them explicit and specifies the direct evidence required to answer them.

Cryogenic electronics forms a rapidly emerging research domain for high-performance and power-efficient computing applications. Incorporating nanomagnetic components in cryogenic circuitry adds highly valuable functionality, facilitating downscaling, reducing energy consumption and introducing time-reversal symmetry breaking. Furthermore, low-temperature environments enhance magnetic stability and switching efficiency at nanoscale dimensions, reinforcing the potential of cryogenic nanomagnets. To fully leverage these opportunities, magnetic control schemes require alternative options to current-based writing, which is the main bottleneck regarding power consumption and downscaling. In this regard, voltage-based gating of the magnetic state could drastically enhance operational efficiency and integration density. In this work, we investigate cryogenic Voltage Control of Magnetism (VCM) in epitaxial SrTiO$_3$/Fe thin film heterostructures on a CMOS compatible Si substrate. We demonstrate and quantify voltage-controlled modifications of the magnetic domain structure, consistent with electric field-controlled magnetic anisotropy at the Fe/SrTiO$_3$ interface. These findings provide a viable material system for the development of next-generation magnetic domain-based devices for classical and quantum computing.

Deep learning performance generally improves with increasing training data, yet this scaling is fundamentally constrained by annotation cost in large-scale dense prediction tasks with long-tailed category distributions, where pixel- or voxel-level annotation is prohibitively expensive. We propose SASS (Stage-Adaptive Sample Selection), a stage-adaptive data-selection framework for pool-based active learning in long-tailed dense prediction. SASS combines three components: label-free self-supervised gradient scoring, prior-guided category rebalancing with validation-driven feedback, and stage-adaptive acquisition aligned with model training dynamics. This design avoids candidate ground-truth masks during gradient scoring while making acquisition responsive to long-tail imbalance and evolving representations. We evaluate SASS on a multimodal 3D medical segmentation testbed comprising over 100,000 samples spanning 108 anatomical structures. SASS recovers 98.3% of full-dataset performance with a 40% training-pool annotation budget, outperforming BADGE by 5.1 percentage points. Moreover, SASS exhibits a statistically supported less-is-more pattern, surpassing full-dataset training at the Hard-group level and, at the structure level, for the pancreas and gallbladder. More broadly, SASS shows that annotation-efficient learning depends not only on which samples are selected, but also on how the annotation budget is distributed across categories and when model-derived scores begin to guide selection.