Papers

Speech brain-computer interfaces (BCIs) aim to restore communication by transforming neural activity related to speech, language, or communicative intent into external outputs such as text, synthesized voice, or avatar control. Recent advances in intracortical and electrocorticographic recording, deep sequence models, and language-model-assisted decoding have enabled rapid progress, including high-performance attempted-speech decoding and increasingly naturalistic speech synthesis. Yet these achievements also reveal that speech BCIs are not simply neural-to-text decoders. They are adaptive clinical systems in which neural representations, recording hardware, decoding architectures, language priors, feedback, and user learning interact over time. Here, we synthesize speech BCI research from a system-level perspective. We first examine the neural substrates of speech and language, emphasizing their hierarchical, distributed, temporally structured, and non-stationary organization. We then examine recording and decoding choices, closed-loop adaptation, evaluation, clinical translation, and ethics. Across these domains, we highlight recurring trade-offs between signal resolution and invasiveness, low-level motor and high-level semantic targets, decoder accuracy and user agency, and language-model fluency and faithful neural evidence. We argue the next generation of speech BCIs should be evaluated not only by offline accuracy, but also by robustness across sessions, calibration burden, latency, uncertainty, usability, and safeguards against unintended decoding. By reframing speech BCIs as adaptive, user-centred systems, we outline the interdisciplinary priorities spanning speech neuroscience, neural engineering, machine learning, clinical practice, and neuroethics needed to move from proof-of-concept decoding toward reliable, expressive, and controllable communication neuroprostheses.

In branch-and-bound (B&B) for mixed-integer linear programming (MILP), branching variable selection critically impacts efficiency. Existing neural branching policies often require GPU inference, while CPU-efficient symbolic expressions lack the representational capacity for complex logic. Large Language Model (LLM)-generated code provides a flexible search space for designing lightweight branching rules with diverse algorithmic logic. To discover effective rules within LLM-based evolutionary frameworks, a core challenge arises: full B&B evaluation on real instances is prohibitively expensive, whereas offline imitation learning suffers from distribution shift. To address this, we introduce a Bi-Fidelity Evolutionary framework (BiFE). It employs low-fidelity imitation scores as a rapid pre-screener and selectively applies high-fidelity on-instance evaluation only to elite candidates, effectively balancing search efficiency with performance reliability. Experiments validate both the search efficiency of BiFE and the competitiveness of its discovered rules, which outperform the SCIP solver and other baselines on CPUs, and even surpass certain GPU-based neural policies.

Knowledge Distillation (KD) trains a smaller-capacity student model to imitate a larger-capacity teacher model by matching output distributions, implicitly assuming the teacher to be a reliable oracle. In large language models (LLMs), this assumption often fails: teacher predictions can exhibit high entropy and hallucinations, causing standard KD to degrade well-calibrated student priors. We propose CaRE-KD, a confidence-gated distillation framework that replaces static objectives with uncertainty-adaptive optimization. CaRE-KD has two components: a token-level loss (CaRE-Divergence) that adaptively switches between Forward and Reverse KL divergence based on teacher–student confidence, and a batch-level epistemic rejection mechanism (Revival) that suppresses updates when the teacher is more uncertain than the student. We provide a gradient-level analysis showing how this dual-granularity design induces a conditional calibration mechanism that prior static divergences cannot reproduce. Empirically, across eight teacher–student pairs and eleven benchmarks spanning instruction following, chat alignment, code generation, and mathematical reasoning, CaRE-KD delivers consistent gains over strong baselines (Skewed-KL, $\alpha$–$\beta$ divergence). Highlights include up to $+3.2$ average ROUGE-L on instruction-following tasks, $+2.1$ pass@1 on MBPP, $+1.7$ accuracy on GSM8k, and $+1.8$ accuracy on CollegeMath over the strongest baseline, with consistent gains in LLM-as-a-judge factuality (up to $+2.5$ per task over Skewed-RKL). Revival further acts as a principled, loss-agnostic plug-in that systematically strengthens existing distillation objectives by filtering epistemically unreliable teacher supervision.

We investigate to what extent the observed nuclear systematics of Short-Range Correlations (SRCs) can be described by the geometry of the underlying proton and neutron density distributions. Building on previous connections between nuclear SRC contacts and one-body densities, we formulate an explicit density-overlap representation in which proton–proton, proton–neutron, and neutron–neutron SRC source terms are constructed from the full spatial density profiles obtained with Energy Density Functionals (EDFs). The framework contains two global parameters describing the overall SRC pair formation strength and the relative contribution of the spin-singlet channel, while the nucleus-dependent evolution is generated by the density-overlap integrals. The framework simultaneously reproduces several independent experimental SRC observables, including proton–proton to proton–neutron pair ratios, relative SRC pair abundances, and proton and neutron high-momentum double ratios. A comparison between relativistic and Skyrme-type EDF families demonstrates the sensitivity of these observables to the underlying density geometry and allows the effect of the complete density profiles to be distinguished from simple radius-based geometrical estimates. Applied to neutron-rich oxygen isotopes, the framework predicts a pronounced evolution of the neutron–neutron pair abundance. Comparison with an independent calculation based on occupied harmonic-oscillator wave functions and an explicit finite-distance criterion supports the predicted isotopic evolution in neutron-rich oxygen isotopes. This agreement identifies the proton–proton to neutron–neutron SRC pair ratio as a promising experimental probe of neutron-skin thicknesses in neutron-rich nuclei and, through their connection to the symmetry energy, of the nuclear equation of state.

Adversarial machine learning has focused mainly on integrity, but availability is an increasingly consequential complement. Latency attacks (also energy-latency attacks) increase inference-time work, energy, or response time, causing deadline misses, throughput collapse, or resource exhaustion in vehicle controllers, interactive services, or battery-powered sensors, sometimes while preserving the nominal prediction. This survey unifies a fragmented literature spanning perception pipelines (including physical attacks on autonomous-driving detection and tracking), input-adaptive neural inference (sponge examples, dynamic networks), and autoregressive and agentic systems (output-length, verbose-image, and reasoning denial-of-service attacks on LLMs, VLMs, mixture-of-experts models, and tool-using agents). We organize attacks by exploited computational bottleneck rather than formulation, separating what makes a computation expensive from how the attacker triggers it; the delivery channel (input, prompt or retrieved content, message, poisoning, or weight tampering) is an orthogonal attribute. Many attacks share one mechanism, intermediate-work amplification, motivating a work-budget defense abstraction; we distinguish caps on the work entering an expensive stage from caps on the results leaving it. We further analyze when a model-level cost increase becomes a system-level availability failure, which depends on critical-path share, slack, existing ceilings, accumulation, resource sharing, and fallback policy, not on the amplification factor alone. We also provide a threat-model taxonomy, consolidated quantitative comparisons, a defense review by control mechanism, and open challenges such as standardized evaluation, physical realizability, and whole-system availability. Companion website: https://github.com/guzonghua/awesome-latency-attacks.

We present an efficient linkage-based model for in-process compartmentalization built on CHERI memory safety, which enables fine-grained compartmentalization of the entire UNIX user-space, scaling to 10K+ compartments on desktop systems. The model's "push-button" compartmentalization along existing library boundaries regularly hosts 500+ compartments per process for large applications such as Chromium, far exceeding the number of concurrently available protection domains supported by other mechanisms (e.g., up to 16 for Intel MPK). Custom policies can further subdivide libraries. Of the thousands of C/C++ programs tested, only the V8 JavaScript engine required source-level adaptation (<300 lines of changed code concerning garbage collection and JIT compilation). We implement the model for CHERI-extended versions of Armv8-A and RISC-V through support in the compiler toolchain and operating system. Case studies illustrate the smooth delegation of memory between compartments, compartment-aware debugging and visualization, as well as extensibility to a complex managed language runtime, demonstrating the benefits of our single-address-space model. We evaluate using multiple processors, including Arm's superscalar Morello and, notably, the first commercial CHERI-enabled RISC-V application core—Codasip's in-order dual-issue X730.

Agents increasingly build on code written by other agents, and they reimplement rather than reuse, growing the codebases later agents must work in. To measure how well agents design libraries for other agents, we introduce LibraryDesignBench, a two-phase benchmark in which an agent implements a full-featured library from a specification that defines required capabilities and potential use cases without prescribing the design. We evaluate the library through the correctness and simplicity of programs written by three user agents from different model families. The benchmark spans 242 expert-validated programming problems across 15 library-design tasks in four languages. On eleven of the fifteen tasks, agent designers reproduce the abstractions of the human-written production library. Downstream agents adopt agent- and human-written libraries alike but underuse them, reimplementing capabilities the library already provides. Our failure analysis finds that downstream agents write extra code mainly because agent-written libraries are rigid or hard to use, not because capabilities are missing. We also experiment with giving designers more prescriptive, agent-first guidance and having them test their library with subagents; this improves downstream scores and yields simpler programs. LibraryDesignBench provides both a testbed for evaluating library-design practices for agent users and an initial design baseline that improves downstream reuse.

A black hole growing adiabatically inside a dark matter halo is expected to develop a dense spike, with a power-law slope between 1.5 and 2.5, depending on redshift and baryonic activity. Direct dynamical tests of this prediction are almost non-existent. Here we show that water megamaser disks offer one; their masers orbit within a parsec of the black hole and are mapped individually with very long baseline interferometry, so the enclosed mass at each radius can be read off the rotation curve. We develop a search method and apply it to eleven megamaser disks with public kinematics. Compared with a black hole alone, a spike is preferred in NGC 1194 and NGC 4258 and moderately improves the fit in several others. Allowing the disks to warp removes this preference in all but NGC 4258 and NGC 6264. No combination of disks, warped or not, allows a spike heavier than about ten per cent of the black hole mass, a sensitivity validated with mock injection tests. Mapping the warps with maser accelerations and detecting fainter inner masers, where a spike and a tilt differ most, can break the remaining degeneracy. Megamaser disks open a window onto dark matter around supermassive black holes, and, importantly, one that enables population statistics.

For a $d$-dimensional probability vector $\mathbf{h}=(h_1,\dots, h_d)$, let $(S^{\mathbf{h}}_n)_{n\geq 0}$ be a nearest-neighbor random walk on $\mathbb{Z}^d$ such that at each step, it moves to one of the two nearest neighbors in the $i$-th dimension with probability $\frac{1}{2} h_i$ ($i=1,\dots, d$). Let $T^{\mathbf{h}}=\inf\{n\geq 1: S^{\mathbf{h}}_n=(0,\dots,0)\}$, the first return time to the origin. For two $d$-dimensional probability vectors $\mathbf{h}'$ and $\mathbf{h}''$ with the former majorizing the latter, we show that $T^{\mathbf{h}'}$ is stochastically smaller than $T^{\mathbf{h}''}$. In particular, the first return time for the $d$-dimensional simple random walk is stochastically larger than $T^{\mathbf{h}}$ for all $d$-dimensional probability vectors $\mathbf{h}$.

HARS (Hybrid Adaptive Reliability-Aware and Syndrome-Aware) is a bit-flipping decoder that combines local channel reliability with parameter selection from the initial syndrome. A bounded check reliability modifies the parity-check weight, while the initial syndrome weight selects the base weight, reliability coefficient, threshold decay factor, and perturbation amplitude for each frame. Computing these quantities at initialization supports synchronous multi-bit updates with a simple iterative datapath. For rate-one-half PEGReg and WiMAX codes over the binary-input additive white Gaussian noise channel, HARS reduces both bit-error rate (BER) and mean iteration count relative to a calibrated SM-NGDBF baseline. The PEGReg BER at 3.25 dB is approximately one-sixth of the comparison value; the WiMAX BER at 2.75 dB is reduced by 59%. A seven-fractional-bit FPGA implementation shares 132 variable-node lanes among four frame contexts. Precomputed check weights, encoded threshold states, and parallel perturbation generation support one group per clock during sustained decoding. On an XC7A200T at 60 MHz, for preloaded 128-frame batches with a ready receiver, measured coded throughputs are 49.7 Mb/s at 2.50 dB and 361.9 Mb/s at 4.00 dB, including initialization and output transfers.

Can a scientific agent distinguish a law it inferred from evidence from one it merely recognizes? We introduce Synthetic Universes, a controlled benchmark that pairs canonical famous worlds with matched twisted twins governed by nearby noncanonical mechanisms. We evaluate each reported law twice: by executing it on held-out continuations and transfer settings, and by independently checking whether it recovers the generating mechanism. In the current checkpoint of a pre-specified 60-cell study, 22 trials were graded and one additional run ended in infrastructure failure. Among 20 twin trials, 8 pass predictive verification while 5 recover the generator. The dissociation is bidirectional: six parsable outputs predict successfully while missing the mechanism, whereas three recover the mechanism but fail predictive rollout. Drag exhibits the first pattern (5/5 predictive pass, 1/5 mechanism recovery); Gravity exhibits the second (1/5 predictive pass, 4/5 mechanism recovery). Because matched famous controls, the corrected identifiability sweep, and the Evidence Ladder remain incomplete, we do not claim a confirmatory causal prior-conflict effect. Instead, the completed runs establish a narrower verification result: predictive adequacy and mechanism recovery are distinct scientific claims and require distinct tests.

We prove two arithmetic nonexistence conditions for tight spherical $5$-designs in dimension $(2m+1)^2-2$. Let $m$ be a positive even integer satisfying $\nu_2(m(m+1))\leq5$, where $\nu_p(a)$ denotes the exponent of the prime $p$ in the positive integer $a$. We show that such a design does not exist if either (i) $m\equiv1\pmod3$ and $\nu_p(m(m+1))\leq1$ for every prime $p\geq7$ with $p\equiv3\pmod4$, or (ii) $m\equiv0\pmod3$, $\nu_3(m(m+1))=1$, and $\nu_p(m(m+1))\leq1$ for every prime $p\geq5$ with $p\not\equiv1\pmod{12}$. Together, these conditions recover the corresponding nonexistence results of Bannai, Munemasa, and Venkov and of Nebe and Venkov, and exclude parameters not covered by either earlier result, including $m=16,40,88,96,100$. The proofs use lattice theory to constrain the discriminant groups of the associated lattices, and derive contradictions through Gauss sums and determinant congruences.

Sparse expert models can distribute traffic evenly while still grouping incompatible training signals within the same experts. We study routing as a gradient-partitioning problem and introduce gradient-aligned routing (GAR), whose load-normalized router objective rewards grouping observations with aligned gradients. On five multi-task text-classification mixtures, we compare GAR with task-loss-only routing, gradient-combination and gradient-conflict methods, and load-balancing losses. With a fully trainable RoBERTa backbone and classification-head experts, GAR has the highest aggregate validation accuracy, 1.07 percentage points above task-loss-only routing. With frozen DeBERTa and Qwen3-1.7B backbones and low-rank adapter experts, it again ranks first, 1.10 points above task-loss-only routing, with better-balanced expert load and higher gradient-mass purity, the share of each expert's gradient-norm mass from its dominant task; the load-balancing losses flatten load further but leave this purity near its task-loss-only level. Top-1 routing, trainable full-parameter feed-forward network (FFN) experts, and a larger backbone also show positive aggregate gains. The results distinguish expert-load balance from gradient-based routing organization and indicate the predictive value of gradient-informed routing in multi-task text classification.

For every $\alpha>1$ we construct a finite set $A\subset\mathbb{D}$ such that the zero-based invariant subspace $I_A$ fails the wandering subspace property. Consequently, the Beurling-type theorem holds on the weighted Bergman space $A^2_\alpha$ if and only if $-1<\alpha\le1$, confirming a conjecture of Shimorin. As a further application, we extend the prescribed-curvature construction of Hedenmalm and Perdomo to all $\alpha>1$, thereby replacing the constant $\alpha_0\approx1.04$ in their result by $1$.

Large language model (LLM) agents repeatedly load reusable content, such as skills, documents, and memory entries, into the current context. Re-encoding this content for every request wastes computation. Position-independent caching (PIC) alleviates this by encoding each artifact independently and reusing its key-value (KV) states at arbitrary positions, but it incurs a quality loss relative to full-context prefill. Existing methods repair this loss by restoring global position IDs or recomputing selected tokens. In this work, we isolate the source of the loss, finding that the positional mismatch has minor effect, and independently cached artifacts retain faithful representations: reading a provided artifact stays largely accurate, and performance degrades only when the model must select among multiple artifacts. Moreover, replacing PIC's attention scores with full-prefill scores recovers performance with the cached KV unchanged, localizing the failure to the attention rather than KV recomputation. Motivated by this, we propose Attuner, a query-side adaptation method that learns to read a frozen artifact cache. Attuner inserts low-rank adapters into the query projections and is trained by distilling full-prefill distribution into the student. It trains fewer than 0.05% of the model parameters and, at inference, requires neither cache recomputation nor a full-context reference. On Qwen3-4B and Qwen3-8B across seven benchmarks covering skills, documents, memory, and code, Attuner substantially outperforms prior PIC baselines in both in-domain and out-of-domain settings, matches full-context prefill quality while providing up to $3.73\times$ speedup.

Fixed-budget adaptation from heterogeneous data sources requires deciding not only how much data to use, but how much exposure each source receives. Size-proportional rules can crowd out small sources, whereas difficulty-only rules can chase noisy estimates or allocate residual budget to nearly saturated pools. We introduce CALIBUDGET, a floor-protected, reliability-aware integer allocator that treats source exposure as an explicit adaptation variable. From small train-internal calibration splits, it combines model need, post-floor availability, and bootstrap stability, then produces exact capacity-respecting quotas without changing the model, objective, or total budget. In a controlled setting combining mathematical and commonsense data, CALIBUDGET improves CommonAvg, FragileAvg, and MacroAvg over validation-error-with-floor, the strongest matched comparator, in all three paired LLaMA-2-7B LoRA+ runs. The respective mean gains are 0.56, 0.46, and 0.41 percentage points (pp). Overall increases by 0.18 pp, whereas MathAvg decreases by 0.20 pp, exposing a coverage-retention boundary rather than a uniform gain. CALIBUDGET changes only 1.14-1.42% of the source budget but improves performance in 15 of 24 comparisons across commonsense tasks and seeds. These results suggest that small changes in source quotas can matter; example-level selection can then determine which examples fill each quota.

Memory-dependent robotic manipulation requires policies to use information that is no longer available in the current observation. Retaining history alone is insufficient: memory must preserve information that supports future actions. One challenge is whether a memory-free foundation model can learn to retain and use historical information from action demonstrations alone, without external memory support. We introduce T$^2$Mem, a framework that develops this capability within a pretrained vision-language-action policy, without external reasoning models or memory-specific annotations. T$^2$Mem uses test-time training to encode observation history into compact fast weights through online self-supervised updates, avoiding repeated processing of the full history. An observation-grounded interface extracts vision-language information for memory formation and supplies retrieved context to the action expert. Action supervision shapes what the memory learns to retain and use, while alternating memory-policy learning gives each component a fixed counterpart during optimization. Across 16 RoboMME tasks, T$^2$Mem improves average success from 17.93% to 56.83% over the memory-free base policy and outperforms the recurrent-memory methods reported in the benchmark, while controlled profiling indicates at least 3x inference speedup over explicit methods. Project website: https://yzliu84.github.io/T2MEM-project/

Phase-field simulations of solid-state dewetting based on polynomial potentials may exhibit spurious bulk-diffusion coarsening, inconsistent with the surface-diffusion-dominated physics. To address this issue, we formulate a degenerate Cahn–Hilliard model with the logarithmic Flory–Huggins potential and dynamic contact line boundary conditions for the film–substrate–vapor triple junction. The logarithmic barrier confines the phase variable to its physical range and suppresses spurious coarsening at the continuum level. We develop a fully discrete, structure-preserving scheme that combines a Lagrange multiplier approach with an upwind finite-volume discretization and rigorously guarantees pointwise boundedness, mass conservation, and energy dissipation without artificial cut-offs or projections. A dimensional-splitting strategy further reduces computational cost while preserving these properties in each one-dimensional sweep. We also derive an explicit estimate of the equilibrium radius contraction caused by spontaneous film shrinking, showing that the logarithmic potential produces weaker spurious shrinkage than the polynomial potential. Numerical experiments confirm the analysis and demonstrate that, for small temperature parameters, the proposed scheme suppresses spurious coarsening and pinch-off while accurately reproducing surface-diffusion-dominated dynamics and relaxation toward equilibrium island structures.

Valley and pseudospin are fundamental degrees of freedom (DOFs) in topological acoustics. However, they have so far been realized mostly in separate systems. Here, we achieve the multiplexing of valley and pseudospin DOFs in one acoustic system. An acoustic metamaterial simultaneously hosting single and double Dirac cones is inversely designed by the developed topology optimization method. Thereafter, under a hard-boundary condition, valley-locked chiral states and pseudospin-locked helical states emerge within two separate frequency windows and can be selected by tuning the operating frequency. Furthermore, their applications for multifrequency acoustic energy enhancement are experimentally demonstrated. Our work provides a strategy to modulate acoustic waves carrying distinct topological DOFs within a single integrated platform, facilitating the development of multiplexed acoustic devices with robustness.

Minimal surfaces bounded by cables or threads arise naturally in tensile architecture: a membrane under uniform tension takes the shape of a minimal surface, and its flexible, inextensible boundary lies along an asymptotic line of constant geodesic curvature. Such configurations have been studied experimentally since the 1960s at the Institute for Lightweight Structures in Stuttgart and more recently realized in gridshells built along networks of asymptotic and geodesic curves. Motivated by this architectural context, we construct two new families of embedded minimal surfaces bounded by finitely many asymptotic arcs of constant curvature, using the Plateau-conjugate method applied to the solution of a partially-free boundary problem for minimal disks that meet the unit sphere orthogonally along the free boundary component. For every integer $m\geq 3$ , we prove the existence of a one-parameter family of embedded minimal annuli, called tensile catenoids, whose boundary components each consist of $m$ asymptotic arcs of constant curvature that meet at cusps. As the parameter varies, the family degenerates from a planar configuration to a union of $m$ minimal disks. For every integer $k\geq 3$, we prove the existence of an embedded minimal surface with the topology of a sphere minus $k$ disks, called a tensile $k$-noid, each of whose $k$ boundary components consists of four asymptotic arcs joined by cusps. Both families are symmetric with respect to a horizontal plane and possess several vertical planes of symmetry. Embeddedness follows from showing that the fundamental piece obtained by conjugation is a graph contained in the region delimited by the planes of symmetry.

Oscillator Potts machines (OPMs) provide a physics-inspired, energy-minimization framework for solving combinatorial optimization problems described by the $q$-state Potts Hamiltonian. Although OPMs may be viewed as multistate extensions of oscillator Ising machines (OIMs), here, we show that they exhibit dynamical properties absent in the binary case. Specifically, we derive a configuration-dependent local-stability condition for a recently proposed multiharmonic OPM formulation and show that configurations with the same Potts energy need not be dynamically equivalent. In particular, for $q\geq4$, permutations of the Potts labels can alter the Jacobian spectrum and, consequently, the regularization strength required to locally stabilize a given Potts configuration. Thus, different phase encodings of the same Potts solution can exhibit different local stability properties despite having identical Potts energies.

DNA labeling has attracted increasing attention in biomedical applications, including molecular imaging, diagnostics, and genomic analysis. In a DNA labeling process, a set of DNA sequence patterns, referred to as labels, is designed according to the requirements of a specific application. For each DNA sequence, the labeling process generates an output sequence that records the positions of the labels. DNA sequences with different labeling outputs can therefore be distinguished through the labeling process. To quantify this capability, the labeling capacity is defined as the exponential growth rate of the maximum number of DNA sequences that can be distinguished through the labeling process as the sequence length tends to infinity [2]. To date, the labeling capacities of several cases in the single-label setting have been determined. In this paper, we formulate the labeling process as a deterministic channel and show that its zero-error capacity is equal to the labeling capacity. For a single label, the corresponding channel can be represented by a star graph. Thus, characterizing the labeling capacity of a single label is equivalent to determining the zero-error capacity of the corresponding star graph. We derive the zero-error capacities of all star graphs, thereby providing a complete characterization of the labeling capacities for all single-label cases. Furthermore, we develop a general method for constructing capacity-achieving codes. These results apply to labeling problems over arbitrary finite alphabets and are not restricted to the DNA alphabet. Finally, for a fixed label length, we exactly characterize the range of achievable labeling capacities and identify all single-label structures that attain the minimum and maximum capacities.

Let $k$ be a field of characteristic zero and let $n\geq 2$ be an integer. We prove that, with $\Z/n\Z$-coefficients, the category of effective motives with modulus is equivalent to the category of effective Voevodsky motives.

Tritium is a naturally occurring radioactive isotope of hydrogen; a low-energy beta emitter with its health significance through internal exposure. The anticipated deployment of fusion energy systems is expected to increase tritium production significantly. Here we present an integrated assessment of tritium generation, release pathways, and resulting environmental concentrations across existing and emerging nuclear technologies. We first compile a comprehensive database of tritium production and releases from monitoring reports and literature. Analysis of light-water-reactor operating data shows that tritium discharge levels and release pathways are governed mainly by plant-specific strategies rather than reactor power or design parameters. This pattern reflects industry operating practices and efforts, with releases remaining well below regulatory limits. Our synthesis further indicates that FLiBe-based (Li2BeF4) systems may increase tritium production and mobility through enhanced permeation of tritiated hydrogen, although mitigation strategies are available to further limit emissions. Environmental transport modeling suggests that, under appropriate siting conditions, resulting tritium concentrations remain below regulatory thresholds. Furthermore, comparative risk analysis indicates that the associated carcinogenic impacts are expected to be lower than those from fossil-fuel-based power generation. Together, these findings demonstrate the importance of a comprehensive monitoring framework for contextualizing and managing routine effluent releases across energy systems.

Accurately learning nonlinear dynamics from a finite-duration experiment requires the efficient collection of informative data. We address this challenge for stochastic controlled nonlinear dynamical systems whose state is observed along a single trajectory. Our goal is to reconstruct the unknown controlled state-increment map over a prescribed compact subset of state-input space. We construct a parametric estimator of the map using fixed nonlinear features, so that the model is nonlinear in the state and input, but linear in the unknown parameters. A Gaussian prior over the parameters yields recursive Bayesian posterior updates as data stream in, enabling online quantification of predictive uncertainty in the reconstructed dynamics over the target set. We formulate an optimal adaptive-design problem over an information state, using a prediction-oriented acquisition criterion based on the mean marginal mutual information between candidate future trajectories and the reconstructed dynamics over the target set. We then approximate the resulting adaptive-design problem by a non-myopic receding-horizon formulation, evaluate its remaining expectation using a scenario-based sample average, and solve the resulting deterministic program with the cross-entropy method, leveraging parallel candidate-scenario evaluations. Numerical experiments on a noisy multistable system demonstrate that the proposed adaptive information-seeking strategy reduces predictive uncertainty and reconstruction error more efficiently than common excitation baselines under comparable experimental constraints.

We examine the spectrum and quantum states of small lattices with cylindrical and toroidal topology subject to a scalar gauge potential that introduces a position dependent phase in the inter-site coupling. Equivalency of gauges assumed in infinite lattices is generally lost due to the periodic boundary conditions, and conditions that restore it are identified. We trace the impact of various system parameters including gauge choice, boundary conditions and inter-site coupling strengths, and an additional axial field. We find gauge dependent appearance of avoided crossings and persistent degeneracies, and we show their impact on the associated eigenstates. Smaller lattices develop prominent gaps in spectral lines associated with edge states, which are suppressed in the thermodynamic limit. Toroidal lattices have counterparts of most of the features observed in cylindrical lattices, but notably they display a transition from localization to delocalization determined by the relation between the field parameter and the number of lattice sites.

Global variational quantum algorithm (VQA) frameworks for solving partial differential equations (PDEs) often rely on a single expressive ansatz over a uniform grid, which becomes structurally inefficient when solutions exhibit spatially heterogeneous complexity such as localized singularities or thin boundary layers. A localized nonsmooth feature can degrade the convergence of the entire global quantum representation, imposing unnecessary circuit depth and amplifying the risk of vanishing gradients in barren plateaus. To overcome this limitation, we propose a block-wise VQA framework for PDEs characterized by spatially heterogeneous solution complexity. The computational domain is decomposed into locally represented quantum subproblems, where grid budgets and ansatz families are dynamically assigned based on a computable difficulty indicator. Artificial block interfaces are coupled through unified jump penalties for state values, derivatives, and physical fluxes, while adaptive reblocking tracks moving rough regions to maintain local accuracy without excessive global qubit overhead. The error analysis rigorously separates spatial discretization, ansatz expressivity, optimization convergence, finite-shot sampling noise, transfer errors from reblocking, and interface-coupling contributions. Reproducible residual-based simulations on representative elliptic, advection–diffusion, and Burgers problems demonstrate lower approximation errors and reduced peak local circuit width compared to global VQA approaches. These results substantiate block-wise quantum resource localization, confirming that high-fidelity PDE solutions can be achieved on near-term quantum devices without requiring a single globally expressive ansatz.

Out-of-distribution (OOD) generalization is a central challenge in scientific machine learning. We study regression problems in which the test distribution differs from the training distribution and ask: under what assumptions on the target function or operator is stable extrapolation possible, and how far beyond the training domain can one extrapolate? Existing theory controls the test error through additive penalties measuring the discrepancy between the training and test distributions. Such guarantees show robustness to small distribution shifts, but can be very pessimistic in comparison to OOD performance observed empirically. We identify classes of holomorphic functions and operators for which the OOD generalization error converges at algebraic rates even in the presence of large distribution shifts. This phenomenon stems from the increasing smoothness of higher-index coordinates, leading to what we term a `blessing of high dimensionality'. For learning with either polynomials, deep neural networks or deep neural operators, we derive explicit rates for arbitrary test measures supported on suitable domains and quantify how the admissible domain depends on the underlying regularity of the function or operator. Our extrapolation guarantees are independent of the test distribution, depending only on its support. We also present a series of numerical experiments across a range of functions and operators that support the main theoretical findings.

Accurate evaluation of $\pi$–$\pi$ stacking interactions requires a description of electron correlation on a small energy scale. Direct variational quantum calculations of the benzene dimer with CAS(28e,20o) require 40 qubits and deep circuits. Here, we propose Divide-and-Conquer Quantum-Selected Configuration Interaction (Deep QSCI), combining the divide-and-conquer idea of Deep VQE with QSCI, and apply it to the sandwich benzene dimer. We perform QSCI for one monomer and construct a reduced local basis by applying particle-number-conserving one-electron excitations to its ground state. Using this basis and the intermonomer interaction Hamiltonian, we build and diagonalize an effective Hamiltonian of the dimer. Since the monomers have the same structure, the QSCI result is reused for both monomers and across intermolecular distances, reducing the required quantum register from 40 to 20 qubits for the monomer CAS(14e,10o). With 6-31G**, the model gives an attractive minimum of -0.91 kcal/mol at 4.0 $\r{A}$, compared with the counterpoise-corrected CCSD(T) value of -1.14 kcal/mol. With cc-pVDZ, a large cancellation appears between the product-state interaction energy and the energy lowering by diagonalization. This may reflect insufficient consistency of the active orbitals and reduced local bases between monomers. Orbital consistency, reduced-space convergence across basis sets, and charge transfer remain key issues for quantitative accuracy. Deep QSCI thus provides a resource-efficient approach to constructing model interaction curves without repeating monomer quantum calculations at each separation.

Legal contracts contain ambiguities that expose enterprises to financial and legal risks. Some ambiguities allow flexible interpretation without triggering disputes, while others lead to significant legal conflicts. This makes identification alone insufficient, and interpretable rationale analysis essential. We propose LAURA, a post-training framework for interpretable ambiguous clause identification. LAURA leverages knowledge distillation with an IRAC-Unlearning prompting technique to transfer knowledge from a teacher LLM to an open-weight student model (<=1B parameters), which is then trained using a joint objective combining classification and rationale generation losses. The framework supports both legal and non-legal stakeholders in making informed decisions about which ambiguities require further attention. Extensive experiments across 7 baselines and 7 open-weight models demonstrate that LAURA with Flan-T5 (250M) delivers state-of-the-art interpretability over all interpretable baselines while matching the identification performance of the best-performing opaque baseline.