Papers

The paper considers a class of optimal set motion problems, where the control determines the velocity of boundary points in the normal direction. Results and counterexamples are given, on the continuous dependence of the minimum cost from the initial configuration. The last sections of the paper study the slicing cost of bounded open sets. For strongly Lipschitz domains, it is proved that the slicing cost varies continuously w.r.t. the Hausdorff distance of the boundaries. We conjecture that the slicing cost for a disc is maximal among all sets with the same area.

Transporting an outcome relationship from a source population to a target population where the outcome is unobserved requires the conditional outcome law to be stable across populations. When it is not (outcome-model drift), the target mean is not point-identified and existing transportability estimators are biased. We study settings, common in cross-cohort educational and biomedical data integration, in which bridge outcomes are recorded in both populations and can be expected to drift alongside the target outcome. The bridge splits the drift into a component it can detect, fixed by a bridge-matching equation, and a residual orthogonal to the bridge that no observed quantity restricts. Because the target is a mean, the residual acts through a single scalar sensitivity parameter $\kappa$. We prove that the identified set has a closed-form irreducible core, whose width is governed by the residual bound and by how much of the outcome the bridge leaves unexplained, and that its center is first-order invariant to the residual. These results hold for any working exponential family, with point identification arising only as the $\kappa=0$ benchmark. We develop a debiased, drift-augmented estimator that is semiparametrically efficient at anchored sensitivity, doubly robust conditional on the bridge-identified drift, and yields rate-robust Imbens-Manski inference for the set. Unlike a bridge-blind sensitivity analysis, which must bound the entire drift channel, the proposed analysis bounds only its bridge-orthogonal part. Simulations confirm that the drift estimator is unbiased under co-drift, that the core width follows its closed form, and that the set fails visibly once the residual bound is exceeded. The motivating setting is the integration of two survey cohorts whose kindergarten mathematics outcome is available in only one.

Sinhala Sandhi splitting recovers the constituent words or morphemes hidden inside a phonologically merged surface form. The task is important for Sinhala NLP because Sandhi obscures lexical boundaries, but no prior published work has established a neural benchmark for Sinhala Sandhi splitting. We present a character-level sequence-to-sequence study based on SandhiLex, using native Sinhala Unicode input and evaluating recurrent encoder-decoder models for affixational and more complex lexicalized, derivational, and etymological Sandhi. The central challenge is the hard subset lexicalized, derivational, and etymological Sandhi, where our best model, a bidirectional LSTM encoder with a unidirectional LSTM decoder, reaches only 68.40% exact-match accuracy (82.08% character-level accuracy), well below the 94.00% achieved on the more regular affixational subset. Ablations show that bidirectional encoding is the largest contributor to performance, while native Sinhala script improves exact match accuracy over romanized input. Qualitative analysis indicates that many errors are near misses involving boundary adjacent characters or plausible but incorrect phonological substitutions. These results establish an empirical baseline for Sinhala Sandhi splitting and identify data scale, Sandhi type conditioning, and attention-based decoding as the main directions for future work.

Long-running LLM agents compress past interactions into persistent memories that may be reused as premises for later tasks. This creates a distinct derivation problem: whether the memory actually follows from what the interaction history supports. Relevant evidence may be scattered across earlier interactions, while compression can introduce relations or event status that the history never established. A valid memory may therefore appear unsupported because its citations omit relevant evidence, while individually supported facts may be composed into a stronger statement the history never established. We characterize this problem through three coupled requirements: (1) Evidence scope; (2) Compositional validity; (3) Admission reliability. We therefore ask whether the interaction history available at write time supports what enters persistent memory. We introduce DerivAudit, a framework for auditing whether a memory is actually supported by the history available when it was written. The audit separates three questions: whether supporting evidence lies beyond writer-provided citations, whether the composed memory introduces unsupported meaning, and how write-time admission decisions affect later memory use. Across two natural memory corpora, audits using broader pre-write history recover support for nearly 60% of memories that appear unsupported from citations alone, while 17-21% remain unsupported after expansion. Yet broader evidence does not by itself make admission reliable: unsupported memories are still frequently admitted across verification models, and evidence expansion alone worsens it on two backbones.

Large language models (LLMs) are increasingly positioned as a force to empower underserved communities, and significant efforts are being made to expand access. Yet, access alone does not equate to meaningful adoption. First, even if a system is accessible, it won't be adopted if users are not willing to adopt it. Second, even if an LLM system is superficially adopted, the heterogeneity of LLM tools means that LLM adoption can be further deepened. Closing this access-adoption gap is critical to ensuring that the full social potential of LLM is not only accessible but fully realised. Drawing on 61 interviews (15 long-form semi-structured interviews with first-generation, low-income college (FGLI) students, 3 non-FGLI students, 3 FGLI program directors, and 40 intercept interviews), this paper examines the access-adoption gap in first-generation, low-income student communities. This paper a) finds that while FGLI students have adopted LLM systems, their depth of LLM tool usage is limited to chatbots (e.g., ChatGPT or Claude) for narrow use cases, and b) identifies barriers limiting their willingness to learn and use (low perceived value, under-estimated self-efficacy, unclear starting point, low peer exposure, and resource constraints). Then, from these findings, the paper derives the four design principles to design a system or an intervention aimed at closing the access-adoption gap in LLM adoption by FGLI students. In doing so, the paper contributes to the field by a) examining the LLM access-adoption gap in the FGLI student community, and b) reframing LLM adoption as a depth gradient across four modes of LLM tool use: basic chatbot interfaces, tool-augmented prebuilt interfaces, agentic development interfaces, and programmatic integration.

Altermagnets' (AMs) non-relativistic spin splitting enables novel states, spintronic and magneto-optical devices, though their optical signatures remain elusive. Here, we report exciton-magnon emission and absorption: optical sidebands from a spin-forbidden dark exciton, a direct consequence of altermagnetic symmetry. Combined optical spectroscopy and first-principles calculations reveal that La$_2$O$_3$Mn$_2$Se$_2$ is an altermagnetic insulator, hosting a strongly bound, spin-forbidden dark exciton and a higher-energy bright exciton. Photoluminescence (PL) and absorption reveal mirror-image sidebands, Stokes-shifted in emission and anti-Stokes-shifted in absorption, symmetric about the dark exciton, whose energy shifts and spectral shapes match the magnon energy scale and density of states measured independently by inelastic neutron scattering. The PL intensity tracks the full equal-time spin-spin correlator, combining static and dynamical contributions, and rules out alternative processes. This directly couples PL to magnetism, with potential for magneto-optical devices. These results establish exciton-magnon spectroscopy as a new route for optically identifying and exploiting AMs.

Unitary process tomography is the act of learning how an unknown unitary gate acts on a quantum system. It is an essential procedure for realizing a functioning quantum computer. As quantum technologies develop, it is important to understand the computational resources needed to conduct unitary process tomography using a variety of types of circuits. In particular, when measuring error in diamond distance, it is known how to achieve optimal query complexity using a sequential circuit architecture, but it had not been known whether a parallel circuit architecture could achieve the same optimal query complexity. In this work, we prove that unitary process tomography with near-optimal query complexity is indeed possible using a parallel circuit by introducing an explicit, novel parallel procedure. This closes the gap between sequential and parallel circuits up to a logarithmic factor, showing that parallel and sequential strategies are nearly equally capable for this task.

Modern AI architectures used to solve visual reasoning tasks typically rely heavily on global connectivity and synchronization. As biological systems demonstrate, though, sophisticated computation can be performed in a more decentralized fashion. In this work, we test the reasoning capabilities of Neural Cellular Automata (NCAs), networks of recurrent cells that use strictly local connectivity and asynchronous updates. NCAs have been extensively studied in artificial life experiments, but it is unclear whether they can perform complex multi-step reasoning. We show that NCAs produce spatio-temporal dynamics capable of solving challenging visual reasoning tasks, including large mazes, Sudoku, and ARC-AGI-1. Furthermore, we provide evidence that NCAs generalize out-of-distribution when running with larger grids, longer rollouts, or parallel trials; and that the latter can be made more efficient via pruning of redundant trajectories. We find that these generalization capabilities depend on training with sample replay and stochastic perturbations, and that stochasticity remains beneficial at test time. Finally, we show that NCAs are robust reasoners capable of dynamically modulating compute to recover efficiently from damage, and that they can scale to solve reasoning in raw pixel space.

Charge-density-wave (CDW) formation in 2H-NbSe$_2$ modifies the low-energy electronic structure and gives rise to collective excitations coupled to the lattice. Here, we investigate bulk 2H-NbSe$_2$ single crystals using far-field terahertz time-domain spectroscopy (THz-TDS) in reflection geometry across the CDW transition at $T_{\mathrm{CDW}} \approx 33$~K. Below $T_{\mathrm{CDW}}$, the THz response shows a pronounced high-frequency feature near 1.5~THz together with longer-lived sub-THz oscillations. Both responses progressively weaken upon warming and are strongly suppressed across the CDW transition, supporting their association with the CDW state. Using time-dependent Ginzburg–Landau simulations, we reproduce the main features of the experimental THz response, associating the high-frequency response mainly with CDW amplitude dynamics and the sub-THz response with defect-pinned phase dynamics. The $\sim1.5$~THz feature lies close to the frequency range reported for the Raman CDW amplitude mode, while coupling to lattice degrees of freedom may also influence its spectral position. We also performed complementary angle-resolved photoemission spectroscopy measurements, which reveal momentum-selective redistribution of near-Fermi-level spectral weight across the transition. Together, these results show that far-field THz spectroscopy provides a sensitive probe of collective CDW dynamics in bulk 2H-NbSe$_2$.

The ultraviolet (UV) spectral slope, $\beta$, serves as a fundamental probe of stellar population properties and interstellar medium conditions in galaxies. At high redshift, a small sample of galaxies have been confirmed to have extremely blue spectral slopes of $\beta<-2.8$, as measured directly from spectroscopy. One explanation for such galaxies is that they are leaking copious amounts of ionizing photons such that these photons are never processed into the nebular continuum and we are observing the bluer intrinsic slopes of the stellar populations. Here, we use a detailed set of radiation hydrodynamics simulations with parsec-scale resolution and non-equilibrium chemistry to elucidate the physical origin of extremely blue UV slopes at high redshift. We show that, while some extremely blue galaxies (EBGs) do indeed have high escape fractions, a second population of EBGs has $f_{\rm esc}<1\%$. In this second population, the ISM gas densities tend to be much lower, and the timescale for ionization fronts to reach their Strömgren radii can be longer than the main-sequence lifetime of massive stars, leading to a lag in nebular emission. Our results demonstrate that extremely blue UV slopes do not uniquely imply high escape fractions, which highlights the importance of time-dependent (non-equilibrium) nebular emission for interpreting the spectra of galaxies in the early Universe.

Loss-cone boundary determination in tokamaks is cast as an active learning problem in which a Bayesian logistic regression model with radial basis function features acts as the learner and the guiding-center orbit integrator acts as the labeling method. The approach is applied to an analytic tokamak equilibrium across a mesh of launch positions, producing a database of loss-probability models with approximate posterior uncertainty, each generated within a prescribed trajectory budget. Once generated, the database supports downstream applications, illustrated here with three examples. First, integrating the probability models against a Maxwellian distribution yields the lost-ion fraction and the energy and parallel momentum carried by lost ions, with posterior uncertainty propagated analytically to each integral. Second, a steady balance between the resulting orbit-loss torque and radial angular-momentum diffusion gives co-current edge rotation whose amplitude is set mainly by the prescribed momentum diffusivity and loss-region renewal rate. Third, a neural network surrogate trained on the database learns the in-domain mapping from launch position to loss-probability image, and transport integrals computed from its predictions closely reproduce the Bayesian reference values on held-out samples from the same mesh.

Thousands of broadband LEO satellites are now in orbit with thousands more on the way, all operating under non-exclusive spectrum allocations across the Ku-, Ka-, and Q/V-bands. To govern this shared spectrum, regulators recently enacted protection constraints that cap instantaneous interference levels and time-averaged throughput degradation. In this article, we review the governing FCC and ITU regulatory frameworks and dissect this interference environment. We examine practical non-cooperative coexistence mechanisms, highlighting how dynamic satellite selection and passive ground sensing circumvent the physical limits of beamforming alone. We then evaluate potential cooperative coordination architectures spanning centralized registries and bilateral interfaces. Finally, we analyze key open challenges: scaling beyond two networks, mitigating feeder link contention from direct-to-cell systems, rerouting traffic over optical inter-satellite links, and modernizing international sharing policies.

Large language models are vulnerable to prompt injection attacks, where third-party adversarial content can hijack the model's behavior. In this paper, we study the role played by the adversarial data's input modality, and identify a systematic asymmetry: multimodal LLMs are more likely to follow adversarial instruction when they appear as text than when the same instruction is delivered through a non-textual channel (e.g., as an image). We hypothesize that this modality gap arises from text-centric instruction tuning, which teaches models to obey textual instructions while treating other modalities mainly as content to parse or describe. We then demonstrate how this gap can be turned into a training-free defense, by rendering all untrusted payloads as typographic images (or audio) before they reach the model. Across ten models and two prompt injection benchmarks (DirectInject and AgentDojo) we show that our defense Pictionary consistently reduces attack success rates even against the strongest adaptive attacks and human red teamers, while largely preserving benign utility. We further show that benign fine-tuning on image-rendered instructions erodes the modality gap, tracing it to the text-centric instruction-tuning distribution.

Learned rotations play an important role in enabling low-bit weight and activation quantization of large language models by smoothing outliers in the activation distribution. State-of-the-art approaches include gradient-based procedures such as SpinQuant and computationally friendlier gradient-free approaches such as DartQuant, but both remain hard to scale to the largest architectures. To address the computational bottlenecks in gradient-free rotation learning, we introduce two ideas for efficiency, (i) a data selection procedure which reduces the required number of calibration data points, and (ii) an exact reduction of the associated optimization on this reduced calibration set. Our data selection procedure exploits the geometric structure of the convex hull of the activations. Using this idea, we show that a carefully selected calibration set with several orders of magnitude fewer activations than state-of-the-art rotation-based methods can match their performance in low-bit quantization settings. Under this extreme data efficiency, the selected activations span an $r$-dimensional subspace with $r<d$, making optimization over a $d\times d$ rotation equivalent to optimizing a $d\times r$ matrix on the Stiefel manifold. We solve this reduced problem using an efficient ADMM algorithm that iteratively employs thin matrix updates at every step, hence the name ThinQuant. For Llama-3-70B with W4A4KV4 quantization, ThinQuant completes the entire rotation calibration in under 12 minutes and achieves a WikiText-2 perplexity of 5.63, compared with 7.55 for DartQuant, which requires 111 minutes. Unlike SpinQuant and DartQuant, ThinQuant also scales to Llama-3.1-405B on a single H200 GPU, completing rotation calibration in just over 2 hours and achieving WikiText-2 perplexity of 2.97 at W4A4, compared with 3.48 for GPTAQ+QuaRoT.

Spatial-temporal (ST) forecasting underpins many real-world systems such as traffic, climate, and energy networks. While existing methods implicitly assume strong spatiotemporal coupling, we observe that real-world ST data exhibits distinct coupling regimes, ranging from temporal-dominated and spatial-dominated to strongly coupled patterns. This mismatch causes current models to suffer from spurious dependencies and degraded performance when one correlation dominates. To overcome this limitation, we aim to dynamically modulate spatial and temporal modeling based on the data's inherent coupling structure. However, three key challenges exist: unknown coupling structure, heterogeneous coupling dynamics, and suboptimal spatial modeling. We propose AdaST, an adaptive ST forecasting framework that tackles these challenges through a decompose-recompose paradigm. AdaST factorizes inputs into components capturing different coupling patterns using heterogeneity-aware experts. Each component is processed by role-aligned modules, and a correlation-informed adaptive recomposer integrates them for final prediction. Extensive experiments confirm that AdaST significantly outperforms state-of-the-art baselines, validating the necessity of an adaptive approach.

Vision-language-action (VLA) policies connect a pretrained vision-language backbone to an action head through a latent interface, but which backbone layers this interface should expose remains unclear. We study single-layer selection and multi-layer fusion for frozen backbones across three pretrained models and two manipulation benchmarks, LIBERO and CALVIN, with three policy-training seeds per configuration. Across three fusion mechanisms and three layer-subset strategies, 47 of 54 configurations underperform the best observed single-layer policy. Our stastical analysis further confirms that fusion's advantage is very limited. However, the best layer varies substantially across backbones and benchmarks, making layer selection consequential and exhaustive policy sweeps expensive. We further derive a reweighting equivalence between the proposed information-bottleneck objectives for action-conditioned InfoNCE and action prediction, motivating InfoNCE as a proxy for layer quality. Empirically, InfoNCE provides the most consistent positive association with policy success among four evaluated proxies. Selecting the layer with the highest InfoNCE score requires 9-33 times less GPU compute than exhaustive policy sweeps and reduces mean selection regret from 17.89 percentage points for deepest-layer selection to 3.71 points across six settings. Its mean regret is close to the 3.28-3.50 points achieved by fixed-layer heuristics optimized retrospectively using all six oracle sweeps, without requiring closed-loop evaluations during selection.

Securing modern AI systems against backdoor attacks remains an open challenge and requires fundamentally principled estimates of the adversary's budget – the poison fraction $\pi$ and trigger strength $\alpha$ needed to construct successful yet stealthy attacks. Motivated by recent empirical evidence that poisoning large language models can require a nearly constant number of malicious samples even as clean datasets grow, we derive an exact closed-form analysis of a quadratic neuron trained on a poisoned Gaussian mixture. We show, perhaps counterintuitively, that the same feature-learning dynamics that make neural networks powerful can also make them more vulnerable to backdoors. Specifically, with clean accuracy preserved to first order, $O(\pi)$, we demonstrate that lazy learning imposes the inverse-square-root scaling $\alpha \propto \pi^{-1/2}$ for a successful attack, while feature learning induces a quadratic detector whose loss margin scales as $O(\alpha^4)$, improving the attack budget to $\alpha \propto \pi^{-1/4}$. Consequently, nonlinear feature learning substantially reduces the trigger strength required at small poison fractions, thereby in a sense making feature learners more backdoor vulnerable. These results provide a theoretical mechanism consistent with large-scale empirical observations and demonstrate that security audits based on linear heuristics can systematically underestimate backdoor vulnerability in the widely adopted feature-learning regimes.

We study how to build more capable general-purpose agents by extending large language models (LLMs) with native typed decision-making. We introduce Dyad, an architecture that augments a pretrained LLM with an environment-conditioned action encoder that embeds each candidate action description in parallel, then scores these embeddings against the LLM's internal state to yield a distribution over typed actions. By factorizing decision-making into representations of the evolving interaction state and environment-specific action semantics, Dyad introduces an inductive bias for learning reusable representations while keeping action scoring efficient even as the action space grows. We investigate two complementary reinforcement learning settings driven by environment interaction. With the LLM frozen, training the action encoder alone achieves consistent gains across four unseen environments, enabling modular adaptation without modifying any LLM parameters. Jointly optimizing both components outperforms conventional RL post-training across diverse interactive tasks and model scales, including a 3.80% average absolute gain on ALFWorld with a 9B model, while improving general knowledge, reasoning, and coding.

Industry applications often demand low-latency classification, yet current large language model (LLM) approaches remain poorly suited for latency-critical applications. Existing prompting and constrained decoding produce verbose, multi-token outputs that require expensive token-by-token generation, while encoder-based models achieve faster inference but sacrifice task flexibility. We propose Koa-decision, a framework for low-latency atomic actions – fast, single-step decisions such as classification, semantic endpointing, Boolean checks, and scoring – formulated as constrained generation with single-token outputs. By introducing atomic label tokens and applying supervised fine-tuning, our method reduces classification to a deterministic one-step decoding problem. Across standard benchmarks, Koa-decision delivers competitive accuracy with consistently low and stable latency. On a production intent-routing benchmark, Koa-decision reaches 85.5% accuracy – competitive with the strongest frontier models (Claude-4.8-Opus, Gemini-Pro-3.1) and ahead of GPT-5 and Gemini-2.5-Pro – while answering in about half a second, several-fold faster than every frontier model (up to ~7.5x at the median) under identical serving conditions. Against the dedicated single-token system Jev/TypeSafe, Koa-decision is competitive on accuracy and faster at the median, while also handling multimodal inputs and multi-label outputs that single-label text systems do not.

The extreme brightness temperatures of fast radio bursts (FRBs) demand a coherent process that converts magnetar burst energy into millisecond radio emission. We propose that FRBs can be produced within magnetar magnetospheres by induced (stimulated) upscattering in a relativistic fireball outflow. In this paper, we focus on induced Compton upscattering in an electron–positron fireball launched from the polar region of a magnetar. Radiation from the X-ray burst accelerates the polar fireball outflow to Lorentz factors of $\Gamma\sim10$–$10^3$. Low-frequency seed waves, such as fast magnetosonic or Alfvén waves, entering the outflow from the side or against its motion undergo induced scattering and are Lorentz-boosted by the relativistic bulk motion from $\nu_0\sim10\mathrm{kHz}$ to $\nu_{\rm obs}\sim\Gamma^2\nu_0\sim\mathrm{GHz}$. Combining the polar fireball dynamics with the linear growth and nonlinear saturation of the induced scattering instability, we derive the FRB luminosity $L_{\rm FRB}^{\rm iso}$ and fractional spectral bandwidth $\Delta\nu/\nu_{\rm c}$. By considering the possible range of pair densities and temperatures, we find that this polar channel can account for the luminosity of Galactic FRB~20200428 and reach luminosities comparable to those of the brightest extragalactic FRBs. If the pair-loaded region is narrower than the relativistic beaming angle $1/\Gamma$, the model also produces a narrow spectrum and predicts $L_{\rm FRB}^{\rm iso}\propto(\Delta\nu/\nu_{\rm c})^2$, a trend consistent with observations of repeating FRBs. The induced upscattering model thus connects magnetar X-ray bursts to FRBs through a coherent emission process grounded in plasma physics and makes observationally testable predictions.

Understanding what limits many-body quantum entanglement is a central problem in physics. Spatial locality has long provided a fundamental mechanism: correlations between a region and its complement must be mediated through a small spatial interface, thereby constraining their entanglement. Here we show that universal constraints on entanglement can persist even when interactions are strongly nonlocal, through collectivity: many weak interactions generate fluctuations controlled by their square-summed, rather than total, strength. We establish this mechanism rigorously for generic gapped Hamiltonians with Kac-normalized power-law interactions $r^{-\alpha}$ on a $D$-dimensional lattice. For arbitrary bipartitions, we prove that the ground-state entanglement scales at most logarithmically with system size for $\alpha<D/2$ and subextensively for $D/2<\alpha<D$, due to suppressed collective fluctuations around individual sites. For spatially regular bipartitions with codimension-one boundaries, we use a renormalization-group construction to extend this suppression to larger length scales, yielding polylogarithmic scaling for $D/2 < \alpha < (D+1) / 2$ and parametrically stronger subvolume bounds for $(D+1)/2<\alpha<D$. We also establish the corresponding optimality results: for arbitrary bipartitions, the logarithmic scaling for $\alpha<D/2$ is optimal and $\alpha=D/2$ marks the optimal threshold for universal logarithmic bounds, while for regular bipartitions the threshold at $\alpha=(D+1)/2$ is likewise optimal for universal polylogarithmic bounds when $D\ge2$. Together, these results reveal collectivity as a fundamental mechanism for constraining many-body entanglement alongside spatial locality.

Circumbinary planets (CBPs) are expected to form in circumbinary discs and migrate towards the disc inner cavity, often near the binary stability boundary. Yet the observed population is not strictly marginally stable: excluding extreme systems, the typical offset is $\Delta \equiv a_{\rm p}/a_{\rm crit}=1.22^{+0.18}_{-0.07}$, suggesting post-formation evolution may shape present-day architectures. We test whether secular evolution of close binaries can contribute to this displacement. Using high-accuracy $N$-body simulations, we follow coplanar CBPs initially placed near $a_{\rm crit}$ while prescribing binary contraction and eccentricity damping along a track that approximately conserves orbital angular momentum. We quantify how the retreat of $a_{\rm crit}(t)$ affects planetary offsets and survival across binary mass ratios. The moving stability boundary filters the near-boundary population, adding a contraction-driven loss component dominated by ejections. Survivors remain quasi-fossilised, with median changes in $a_{\rm p}$ and $e_{\rm p}$ close to zero and semimajor-axis drifts of only a few per cent. For all valid synthetic survivors, the final offset has median $\Delta_f=1.86^{+0.18}{-0.26}$. Applying the same damping prescription to the observed sample shifts the median from $\Delta{\rm now}=1.22^{+0.18}{-0.07}$ to $\Delta{\rm damp}=1.44^{+0.32}_{-0.10}$. Near-equal-mass binaries undergo stronger filtering, with enhanced ejections and collisions. Binary tidal evolution can therefore shape non-marginal circumbinary architectures. By removing planets close to marginal stability while leaving survivors nearly fossilised, the retreat of $a_{\rm crit}$ can widen the inner gap and increase offsets from the present-day stability boundary. This links binary orbital evolution, dynamical filtering, and transit detectability in a multi-channel interpretation of the CBP desert.

External memory interfaces (such as LPDDR5/4 and DDR5) are essential components of modern mobile, cloud, and enterprise computing systems. While memory manufacturers focus on physical DRAM die development, the vast majority of semiconductor firms design custom application ASICs that require a dedicated Memory Controller to interface with these standardized external memories. Consequently, pre-silicon design verification of the Memory Controller RTL – acting as the Device Under Test (DUT) against a third-party DRAM Verification IP (VIP) – is a ubiquitous and critical challenge across the global semiconductor industry. This article presents an automated, pre-silicon configuration and closed-loop initialization framework for Memory Controller verification. The proposed solution parses JEDEC timing and configuration parameters directly from the DRAM VIP's database files (such as Denali SOMA files) to configure the Memory Controller DUT's registers, while a custom Mode Register Register Abstraction Layer (MR-RAL) tracks volatile DRAM VIP states in real-time. Furthermore, a dynamic, re-compilation-free PHY initialization flow randomizes interface parameters directly in the testbench, executes on-the-fly configuration generation via system calls, and parses the resulting register write sequences at runtime. By integrating these automated methodologies into pre-silicon verification flows, engineers can eliminate setup overhead, prevent false protocol violations, and enable comprehensive randomized testing of complex PHY and Memory Controller configurations.

We present a statistical evidence of a temporal correlation between ionospheric storm-enhanced density (SED) and plasmaspheric plumes during magnetic storms. We identified SED plumes in 75 storms between 2010-2024 with sufficient total electron content (TEC) coverage. We also identified plasmaspheric plumes during these storms using a plasmapause asymmetry (PPA) index derived from the machine-learning-based DEN3D model. We found that all SED plumes coincided with plasmaspheric plumes. The start and end times of both phenomena exhibit near-zero lag within a half-hour delay, confirming their synchronized evolution. For the first time, superposed epoch analyses reveal that plume onsets align with solar wind driving and geomagnetic activity, including peaks in the electric field, coupling function, auroral electrojet indices, and the asymmetric ring current index, all of which facilitate plume formation. This work provides timely new insights into magnetosphere-ionosphere coupling processes and demonstrates the utility of machine-learning-based diagnostics for space physics/weather.

In a recent paper arXiv:2503.20727, Ichino and Prasanna proved a supercongruence for binomial coefficients and related it to representations of $\mathrm{GL}_2$ over $p$-adic rings. We prove a $q$-analog of this supercongruence using the $q$-Lucas theorem, from which a strengthening of the main result of loc. cit. can easily be deduced. Our proof also substantially simplifies the original argument of arXiv:2503.20727.

Tabular foundation models using in-context learning have recently surpassed gradient-boosted trees on predictive tabular tasks. However, recent mechanistic insights suggest that parameters in these models are largely redundant. We introduce LoopICL, a looped transformer whose core design decouples parameter count from computational depth. LoopICL consists of a single block, processing data through two coupled streams: a cell stream capturing per-cell feature representations and a row stream capturing in-context example representations, jointly refined through within-column and cross-column attention. During pre-training, we vary loop counts, allowing the block to be unrolled for a varying number of iterations at test-time and use a learned exit-gate to automatically exit. In its standard setting, LoopICL performs competitively with TabICLv2 on TabArena and TALENT at the same computational cost (FLOPs), while using nearly 90% fewer parameters. Furthermore, its recurrent design enables users to also trade off inference cost and performance, providing a resource-aware TFM.

While pretrained robotic policies exhibit impressive capabilities in controlled environments, unobserved physical properties and dynamics require these policies to rapidly adapt during deployment. Existing test-time adaptation methods typically rely on sparse scalar rewards, failing to exploit the rich geometric and dynamic feedback from the environment during physical interaction. To address this challenge, we propose SCOUT, a dynamics-aware meta-learning framework that enables manipulation policies to rapidly adapt by continuously revising their internal beliefs about environment dynamics. Our approach couples an action-prediction policy with a forward dynamics model via a shared belief latent space. During meta-training, an inner loop updates this shared belief latent by minimizing the dynamics prediction error against the observed action outcome, while the outer loop optimizes the network for action selection. At deployment, this structure allows the agent to infer and adapt to unknown physical dynamics on the fly. By updating its latent belief based on action-outcome mismatches, the policy automatically adapts without risking catastrophic forgetting. We demonstrate that SCOUT significantly accelerates online adaptation across simulated manipulation benchmarks and achieves robust sim-to-real transfer in the real world. Project webiste can be found here: https://liy1shu.github.io/SCOUT/

We study the fundamental limits of transferability in algebraic signal processing through homomorphisms between algebraic signal models. Homomorphisms are linear maps between the signal spaces of two models that commute with filtering, so filtering a signal and transferring it across domains can be done in either order. The existence of such maps is governed entirely by coincidences among the filtered eigenvalues of the two models' shift operators, but existence alone is insufficient: the space of homomorphisms always contains trivial elements that destroy all information. We introduce the spectral transfer efficiency $\eta(\theta)\in[0,1]$ to quantify information-preserving quality, prove that every homomorphism decomposes into unconstrained blocks over coincidence classes, derive the dimension of the homomorphism space, and characterize exactly when lossless transfer is achievable. Beyond normal shift operators, we quantify a departure-from-normality penalty and show how filter derivatives can repair spectral defectiveness. The theory yields concrete consequences in three settings: for sampling, eigenvalue interlacing converts transferability under subsampling into an explicit filter design constraint; for compressed sensing, $\eta(\theta)$ controls the restricted isometry constant and the coherence of the resulting measurements, and recovery decouples across coincidence classes; and for machine learning, spectral aliasing emerges as the controlled symmetry breaking that makes transfer between mismatched domains possible at all.

Let $f$ be a Hénon map of $\mathbb C^2$. We provide quantitative versions of several results involving dynamical objects associated to $f$. Our results can be interpreted as a quantitative version of Pesin theory with geometric control. As an application we show that the perdiodic points of period $n$ of $f$ equidistribute towards its equilibrium measure exponentially fast as $n$ tends to infinity.

Replacing one LLM agent with a collaborating team can raise accuracy, but whether scaling the team helps, and which architecture to scale, is unclear. Sweeping eight agent orchestration architectures across five instruction-tuned 7-9B models, five short-answer benchmarks, and an executable-code benchmark up to 30 calls, we find that the returns to team scaling are sharply task-dependent: from three to thirty calls accuracy rises by up to 17 points on the two arithmetic word-problem benchmarks (GSM8K, GSMHard) but by at most four on ARC, GPQA, and MMLU, for every architecture, a split the usual task-averaged number conceals. Proposer-Critic captures the arithmetic gains, scaling steepest and, in aggregate, surpassing every other architecture at the largest budget (item-clustered intervals exclude zero), though it ranks among the weakest elsewhere, and no architecture wins across tasks. We explain these trajectories with an exact generate-transform decomposition. Partitioning any workflow into proposal coverage and a downstream transform, any accuracy change splits exactly into an extensive coverage dividend and an intensive transformation change. The decomposition diagnoses each task: arithmetic offers coverage headroom that a critic-guided transform converts, whereas the multiple-choice benchmarks either saturate in coverage or fail to convert it, and on open-ended code generative recovery nearly vanishes so accuracy tracks coverage. At equal call budgets token cost still varies 2.1x. Extra calls therefore create candidate opportunity that only some architectures, on some tasks, convert. Team scaling is a task- and architecture-specific bet, not a uniform lever.