Papers

We consider quantum circuits consisting of $d$ layers of nearest-neighbor two-qubit gates acting on $n$ qubits arranged on a line, where every qubit is independently depolarized with a constant probability before each layer. We describe a randomized parallel algorithm which samples from the output distribution of any such circuit to within total variation error $\delta$, with parallel runtime $2^{O(d)}\log\log(n/\delta)$ and $n 2^{O(d)}$ elementary real-arithmetic operations. Without noise, the same approach gives an exact sampler with parallel runtime $O(\log n)$. For constant depth, we further show that the input/output behavior of the noisy quantum circuit is reproduced up to a constant error by a randomized $\mathsf{AC}^0$-circuit, that is, a Boolean circuit of polynomial size and constant depth with unbounded fan-in AND and OR gates and NOT gates. Consequently, every relation problem solved by a noisy constant-depth quantum circuit in one dimension is also solved, with essentially the same success probability, by a randomized $\mathsf{AC}^0$-circuit. This rules out, for noisy circuits in one dimension, the unconditional quantum advantage established for noisy shallow circuits in two and three dimensions, where polynomial-size classical circuits over the same gate set require depth $\Omega(\log n/\log\log n)$. Our algorithm exploits the fact that depolarizing noise cuts a one-dimensional circuit into independent pieces of logarithmic width, each of which can be sampled exactly in parallel.

Interpreting properties of LLM-generated data is important for understanding its utility and limitations across learning tasks. In this work, we characterize synthetic data through sample-level learnability, studying variation among LLM families and scales, alongside human-written data as a reference. We first generate synthetic datasets spanning single- and multi-label classification, labeling, and tree prediction tasks. We then derive empirical data distributions from encoder training dynamics for both machine and organic data, and estimate the robustness of these distributions across encoders. Finally, we evaluate how data selection strategies based on these learnability signals affect both data sources differently.

Existing full-duplex speech benchmarks cover only subsets of real-time interaction behaviors, often under limited contextual conditions. We introduce DuplexAct-Bench, a bilingual benchmark that systematically covers six complementary behaviors, from interruption and yielding to proactive initiation, active silence, and backchanneling, across Pre-session, In-session, and No-explicit conditions. Across 1,290 English and Chinese streaming trials, we evaluate 12 full-duplex speech systems on both Timing and Content. Results reveal substantial variation across behaviors, conditions, and systems, as well as frequent mismatches between semantic quality and behavioral timing. These findings show that current systems remain far from robustly managing when, whether, and how to participate as real-time interaction unfolds. Project page: https://alitaxky.icu/DuplexAct-Bench/

Time series foundation models (TSFMs) commonly adapt to new data by attaching a single trainable head to a frozen backbone, a one-size-fits-all setup that underfits heterogeneous regimes. Replacing the head with a mixture of experts is the standard upgrade, but on instance-normalized backbones (the dominant TSFM design class) it fails: routing entropy collapses to zero and one expert absorbs every input, a failure we call normalization-induced routing collapse. Standard MoE rescue mechanisms do not repair it, because the cause is in the router's input, not its optimization. Pre-encoder normalization strips the statistics a router would need to tell regimes apart. A mutual-information decomposition makes this precise and yields a signal-ratio that, computed before training, predicts dataset vulnerability (Spearman $\rho = -0.88$). Eight causal controls, including a vision-modality replication, isolate instance normalization as the cause. The prescription is a minimal causal intervention: Raw-Routed Mixture of Adapters (RR-MoA), which routes on the raw, pre-normalization input. Under a strictly frozen backbone, RR-MoA wins 54/54 comparisons against the strongest fixed adapter and significantly outperforms LoRA, TRACE, AdaMix, and full fine-tuning. The effect generalizes across six backbones and an imputation task. Frozen RR-MoA also beats full fine-tuning by 12-79% (the Frozen Paradox); two architecturally distinct variants confirm the principle generalizes beyond this specific router.

We study bistable pulsating waves for reaction-diffusion equations with general periodic advection in arbitrary space dimension, allowing the diffusion matrix to be nonsymmetric. Assuming that the homogenized equation admits a traveling wave with nonzero speed in a given direction, we construct moving pulsating waves for all sufficiently small spatial periods $L$ and prove their convergence to the homogenized wave as $L\to0^+$. The existence range and convergence are uniform in the propagation direction when the homogenized speeds never vanish. We also prove uniqueness of the wave speed for arbitrary periods, profile uniqueness for moving waves, and stationary-wave uniqueness in the standing case under continuity of the competing profile. For spatially homogeneous reactions, we derive the expansion $c_L=c_0+Lc_1+O(L^2)$ and an explicit formula for $c_1$. Examples with constant diffusion and zero-mean periodic advection show that $c_1$ can have either sign, demonstrating that the heterogeneity in advection may accelerate or decelerate propagation relative to the homogenized limit and that the general $O(L)$ speed estimate is sharp.

Slender flexible structures such as mooring lines, aquaculture netting, and aquatic vegetation are too thin to be resolved economically by a body-fitted mesh, yet they exchange momentum with the flow in both directions. This paper presents a finite volume actuator-line formulation for the fluid–structure interaction of such structures, in which a geometrically exact Simo–Reissner beam, discretised by cell-centred finite volumes, is coupled two-way to an incompressible finite volume flow solver through conservative, parallel-consistent transfer operators. The fluid velocity is sampled upstream of each beam control volume, a quasi-steady Morison-type drag closure evaluates the line force, and the equal and opposite reaction is projected into the fluid momentum equation through a normalised Gaussian kernel that conserves force. The two subsystems share a discretisation, time loop, and domain decomposition, so no external structural code is required, and the OpenFOAM implementation is released publicly. The accuracy of the drag-only closure and the conditions over which it holds are quantified against two channel benchmarks and a wave-flume experiment, with all drag coefficients fixed in advance. In a steady confined channel flow the tip deflection of a wall-mounted cantilever falls within the range of published continuum solutions. For a flexible leaflet in sinusoidal channel flow the peak tip excursion departs from the reference at the longer forcing period. The leaflet then moves further relative to the surrounding fluid, so added-mass and history forces, which the closure omits, become important. In the wave flume the simulated stem reproduces the measured mode shape, intra-wave asymmetry, and base force, and recovers 71–83% of the oscillation range. A quasi-static reconstruction test places the remaining shortfall in omitted fluid inertia rather than in the structural model.

Generative artificial intelligence can contribute to the representations, alternatives and evaluations through which executive judgements are formed. Existing research already explains important aspects of hybrid cognition, reliance, managerial agency and accountability. This article develops a narrower process challenge: locally competent human and AI contributions can still culminate in an inadequately warranted organisational commitment when established operational assumptions, uncertainties or dependencies are lost or transformed before authorisation. It proposes qualification attrition as the process through which such epistemically consequential content is weakened across successive transformations, and a formation–authorisation gap as the resulting discrepancy between the grounds required for warranted commitment and those that remain accessible, intelligible and challengeable to the authorising role. Executive judgement governance is defined through four dimensions: boundary setting, interpretive challenge, reliance calibration, and authorisation with answerability. Four propositions connect transformation, qualification attrition and governance arrangements to epistemic risk, accountability risk and judgement quality. The contribution is an integrative process theory of how distributed judgement formation can fail as a composition even where individual contributions, local reliance decisions and formal accountability arrangements appear competent.

Online reinforcement learning has been extended to flow matching for diffusion model (DM) image generation. However, this paradigm faces three limitations: (1) Window selection. Existing methods manually set the stochastic differential equation (SDE) sampling window, i.e., the denoising steps where exploration noise is injected. We instead determine it from each model's denoising trajectory. (2) Reward saturation. Current methods rely on scoring models trained on human annotations; we find that such scores are extremely high and nearly indistinguishable on the latest SOTA open-source DMs, making advantage estimation largely ineffective. (3) Sample inefficiency. A single scalar reward collapses different failure modes into almost identical scores, leaving minimal gradient guidance for targeted improvement. To address these issues, we propose CAST (Causal Advantage-Structured Training), an RL fine-tuning method for pretrained DMs, which (1) identifies the denoising step at which each model fixes the objects and their spatial arrangement in the image and uses that timing to set the SDE window, (2) decomposes each prompt via Causal Scene Graphs (CSG) into verifiable-atoms, i.e., minimal semantic units such as an object, count, attribute, or spatial relation that can each be checked independently, and rewards each atom separately, and (3) projects the signed atom-level advantages into pixel space through teacher-forced attention and uses them to spatially weight the SDE policy objective. We fine-tune two of the strongest open-source DMs, FLUX.2-dev and Qwen-Image-2512, with CAST, and evaluate them on GenEval 2, a compositional benchmark, and on Qwen-Image-Bench for overall quality. Within almost the same training budget, CAST's improvement over the base model on the most challenging GenEval 2 prompts is up to 3.07x that of Flow-GRPO, while overall generation quality also improves.

We compare the instance-wise, finite-sample risks of monotone spectral filters for linear regression, a broad class of estimators including principal component regression (PCR), gradient descent (GD), and ridge regression. We show that PCR dominates all monotone spectral filters: compared to any such filter, the risk of optimally tuned PCR is no bigger by a constant factor for all problems. Furthermore, the dominance is strong if the filter is separated from step functions (e.g., GD and ridge): there exist problem instances for which the risk of PCR is smaller by a polynomial factor in sample size dependence. Our comparison results show that PCR is optimal and thus admissible among monotone filters, significantly extending Wu et al. (2026)'s result that GD strongly dominates ridge. From a technical perspective, we establish new upper and lower bounds for general spectral filters, which are instance-wise sharp when specialized to ridge or GD, recovering or improving the best-known bounds.

Active spinners inject angular momentum into their surroundings without persistent translation, providing a route to chiral active matter distinct from self-propelled particles moving along curved trajectories. Recent studies have shown that spinner fluids can self-organize into rotating clusters with circulating edge currents characteristic of odd-viscous fluids. Whether angular-momentum injection alone can transmit chirality and edge-current dynamics to ordinary passive matter - and thus serve as a generic mechanism for imparting chiral functionality to undriven material components - remains an open question. Here, we show experimentally and numerically that purely rotational magnetic spinners can transfer chirality and edge-current dynamics to passive colloidal matter. At high passive-particle fraction, active spinners assemble into rotating clusters surrounded by passive matter. Remarkably, upon changing composition and spinner attraction, this organization turns inside out: passive colloids form rotating clusters surrounded by a chiral spinner fluid. These passive rotors exhibit edge currents and the same angular-velocity scaling, $|\Omega|\sim R^{-2}$, as their active counterparts. A minimal model shows that spinner-mediated transverse and non-reciprocal interactions are sufficient to reproduce these complementary chiral states and their boundary-driven dynamics. Our results establish angular-momentum injection by active spinners as a route to endow otherwise passive matter with chirality, collective rotation and boundary transport.

In integrated sensing and communication (ISAC) systems involving multiple operators, unmanned aerial vehicle (UAV) tracking can benefit from sensing at multiple base stations, while raw sensing records, channel information, and transmit decisions remain local to each operator. The resulting problem is to obtain a reliable UAV prediction from different track estimates and use it to guide transmission without centralizing their data or control variables. To resolve this, we propose a two-timescale framework comprising slow-timescale motion-uncertainty calibration and online track fusion with predictive beamforming. Local radar innovations calibrate the uncertainty assigned by each tracker to unmodeled UAV motion, with only the resulting parameters shared across operators. During online tracking, covariance intersection (CI) combines local estimates without requiring knowledge of correlations between their estimation errors. The fused prediction then guides each operator's transmission design subject to communication QoS, power, and tracking constraints. We further establish exact rank one recovery for the communication covariances and finite interval bounds on local tracking uncertainty. Simulations show improved consistency of tracking uncertainty under substantial motion model mismatch. CI avoids overconfidence under correlated errors, while using the current fused prediction reduces RMSE by about 31% relative to transmission designed for communication alone, with less than 2% additional power. Ablation further shows that the power difference between current and outdated predictions is driven mainly by predictive covariance rather than predictive mean.

Large language models (LLMs) encode substantial latent geographic knowledge, yet they reason poorly over space and are unreliable when queried from coordinates alone. Useful behaviour emerges only when structured spatial context is supplied in the prompt. This raises a question geographic evaluation has left unexamined: once the right context is supplied, does the model reason from it, or override it with its own parametric recall? We take up this question with an open pipeline for network-based spatial context retriev-al. In it, the surroundings of a selected point are defined by the pedestrian street network, the area actually reachable on foot. Using only open data and open-weight models, the pipeline retrieves features from OpenStreetMap and the GHS-POP population grid, computes indicators over the network catchment in code, and injects them as a compact spatial brief. On this basis we build a faithfulness benchmark. It labels every claim a model makes by its source (grounded in the brief, or drawn from training knowledge) and its correctness, and it probes each case with a planted false premise that the brief refutes. We evaluate sixteen open-weight model configurations across three families (Qwen, Gemma and Llama, with Gemma in two generations), four size classes and, where available, both thinking and non-thinking modes, on three con-trasting cities, resampling every case over ten seeds. The results show that resistance to the planted premise varies more strongly by model family and generation than by scale, while brief-reading competence forms a partly separate dimension. These behaviours are not captured by conventional world-correctness scores or single-shot evaluation. We release the implementation, spatial briefs, model outputs, and claim-level labels as a reproducible workflow at github.com/perezjoan/NSCR-LLM.

Reinforcement learning with a verifiable reward (RLVR) offers a scalable approach to training language-model agents, yet sparse outcome rewards can leave early training with little signal for policy improvement. We identify an On-Policy Acceleration Phenomenon: in our main comparisons, RLVR initialized with on-policy distillation reaches high performance earlier in training and achieves both higher average performance during subsequent RLVR and higher final performance than the alternative baselines. Motivated by this observation, we study On-Policy Warmup (OPW), a teacher-guided stage in which the student trains with teacher supervision on its own interaction trajectories before transitioning to RLVR. Unlike imitation on fixed teacher-generated trajectories, OPW targets states induced by the student's own decisions, including imperfect actions and recovery situations. We provide a theoretical explanation by connecting on-policy reverse-KL distillation to trajectory-level distribution matching. Under a competent teacher and sufficiently small population distillation loss, this connection yields a lower bound on initial verifier success and a corresponding bound on reward-discovery complexity. For group-relative RLVR, we further characterize when increased success probability produces more reward-informative groups. Together, our findings support on-policy distillation as an effective warmup for agentic RLVR and identify initial reward discovery as a mechanism that can contribute to the observed acceleration.

We report the observation of a photomagnetodiode effect with a giant rectification ratio of 10 at a magnetic field of 0.1 T. The effect consists of a diode-like dependence of the photocurrent on the applied voltage that emerges in an external magnetic field. Such a pronounced nonreciprocity is observed in high-quality VPE-grown GaAs samples at low temperatures at interband photoexcitation. The photoconductivity is invariant upon reversing the polarity of both electric and magnetic fields. Similar behavior with even higher asymmetry ratio is observed in photoluminescence. The findings are well described by the theory of ambipolar drift of carriers in electric and magnetic fields together with fast surface recombination.

We study quiver varieties associated with orthogonal quivers and determine the Kleinian surface singularities arising from affine Dynkin types. In the classical setting, affine Dynkin quivers play a distinguished role through Kronheimer's construction of Kleinian singularities and its reformulation by Nakajima in terms of quiver varieties. Using a graphical calculus in the preprojective algebra of the quiver, we determine which Kleinian surface singularities arise from orthogonal quivers whose underlying graph is an affine Dynkin diagram.

Everyone seems to be thinking about so-called artificial intelligence (AI) these days. As scientists, we understand that we're not about to have a conscious computer intelligence taking over the world, but we're nevertheless concerned about how "large language models" (LLMs) and AI "agents" are impacting the way that we do science. This concern has generated several recent essays about the effects of AI on research, including in my own field of astronomy and astrophysics. These opinion pieces tend to miss the biggest current concern for those whose jobs combine teaching with research - although the threat to future scientific research as we know it is probably a bit concerning, the effect on teaching is an existential crisis right now! We need to hang on to the parts of education that we think are at its core - so that students coming out of degree programmes are actually able to think for themselves, "like an astrophysicist". This means emphasising to students that they should use AI tools to support their education and not to replace it. One way of explaining the importance of this principle is to point out that students should be training their own neural networks.

We investigate the potential of future observations of the Dark Ages 21 cm power spectrum to probe the scale dependence of the primordial power spectrum and distinguish between inflationary models. We find that the Dark Ages 21 cm power spectrum can serve as a powerful probe of inflationary physics on scales beyond the reach of current CMB observations. While models predicting nearly identical running spectral indices may remain difficult to distinguish even with future 21 cm observations, models with unconventional scale dependence can be stringently tested. We further show that combining 21 cm observations with CMB data can significantly improve the constraints, even for a moderate observational configuration. We examine the impact of the choice of pivot scale on the resulting constraints.

We develop a sharp-interface method for freely moving solid particles undergoing pure or binary melting. A Level-Set representation preserves the particle geometry during long-distance translation and rotation, while an embedded-boundary formulation imposes hydrodynamic, thermal and solutal conditions sharply on a Eulerian Cartesian mesh. Liquid temperature and solutal concentration are advanced with a conservative moving-cut-cell discretization, whereas the solid temperature is transported through a geometry-consistent conservative tracer. Hydrodynamic loads, rigid-body motion and phase change are coupled iteratively within each physical time step. The method is validated progressively for rigid-body motion, prescribed-motion pure and binary melting, and fully coupled melting with free particle motion. It reproduces established reference solutions and captures wake-induced rotation of a melting sphere and the simultaneous translation, rotation and thermosolutal melting of a three-dimensional oblate ellipsoid. As a final application, simulations of a floating ice disc reveal how thermoconvective instability, geometric confinement and melting jointly determine its spontaneous motion, with phase change feeding back on the instability through the evolving particle geometry. Additional tests establish the importance of long-time geometrical fidelity, local scalar conservation and geometry-consistent solid-temperature transport. The resulting framework provides a sharp and conservative approach for strongly coupled moving-body phase-change problems.

Analogues of the Sylow, Hall and Schur-Zassenhaus theorems for finite groups have recently been generalised to finite skew braces, providing a powerful framework for their structural analysis. Building on this foundation, the natural next step of skew braces is the study of Carter sub-skew braces, namely, centrally nilpotent, self-idealising sub-skew braces. However, while every finite soluble group possesses Carter subgroups, not every finite soluble skew brace possesses Carter sub-skew braces. We introduce Carter sub-skew braces, analyse their structural limitations and establish Carter-like theorems. We prove that finite soluble skew braces have Carter sub-skew braces if every sub-skew brace of prime-power order is a centrally nilpotent left ideal. We also prove their existence within finite multipermutational skew braces.

Uncertainty Quantification (UQ) is a key requirement for trustworthy AI in high-stakes medical image analysis. In this work, we evaluate UQ in a multi-task Deep Learning framework for MRI-based glioma diagnosis that performs tumor segmentation and predicts IDH mutation status, 1p/19q co-deletion status, and tumor grade. Monte Carlo Dropout (MCD) is used for a detailed task-aware analysis of predictive, aleatoric, and epistemic uncertainty. We assess MC sample convergence, calibration, error detection, selective prediction, associations with segmentation performance, and the effect of voxel-wise uncertainty aggregation on case-level reliability. We also compare MCD with Deep Ensembles (DE) and Monte Carlo Deep Ensembles (MCDE), examine interactions between segmentation quality and classification, and evaluate a composite trust score integrating segmentation and classification uncertainty. Across tasks, uncertainty estimates supported meaningful error detection, while calibration depended on the dropout rate, with moderate rates yielding the most reliable probabilities. Uncertainty decomposition provided task-dependent interpretability but did not consistently improve error detection over predictive uncertainty alone. DE and MCDE showed comparable operational utility, with no method consistently dominating across tasks and metrics. The composite trust score did not consistently outperform classification uncertainty for selective prediction. Overall, our results provide a task-aware evaluation strategy and practical guidance for the development of trustworthy AI for glioma diagnosis.

The long-term evolution of neutron star crustal magnetic fields through Ohmic decay is governed by the transport properties of the crust. In this work, we demonstrate that the depth at which the magnetic flux is anchored plays a significant role in the dissipation timescale. We study the evolution of the magnetic field at the surface of the neutron star due to the crustal field evolution, under the assumption of Ohmic diffusion in a curved spacetime. We employ a cumulative distribution function to parameterise the initial current distribution and explore the effects of varying the equation of state and crustal impurity. We compare results across three neutron star models representing varying degrees of compactness. We find that magnetic fields anchored deeper within the inner crust experience negligible decay over timescales up to $10^{10}$ years. Significant field evolution, necessary to explain the deficit of high-field pulsars at advanced ages, is possible if the magnetic currents are confined to the shallow outer crust. This result holds true regardless of the impurity concentration or the specific geometric thickness of the crust determined by the equation of state.

Detecting missing components on printed circuit boards (PCBs) differs fundamentally from conventional object detection, as the model must localize components that are not present. We introduce PCB-MC, a curated dataset for missing component detection with footprint level annotations built on top of the RF100 dataset. The dataset contains 197 distinct board types, each corresponding to a unique PCB design, with multiple augmented samples per type. We also provide benchmark results on PCB-MC by evaluating a diverse set of supervised and unsupervised methods. To ensure fair evaluation, we propose board type aware cross validation splits that prevent layout leakage between training and test sets. Supervised models showcase high false negative rates on unseen board designs, and unsupervised anomaly detection methods fail entirely due to the lack of spatial alignment with a board specific reference. These results confirm that missing component detection on diverse PCB layouts remains an open challenge. We release PCB-MC and all training protocols to support reproducible research on structural absence detection in industrial inspection.

Human decision-making reflects not only material incentives, but also concern for others' gains and the ability to understand others. However, the extent to which sensitivity to fairness interacts with aspects of social cognition and affects sequential social interactions requires further understanding. Here, we address this question through a behavioural experiment on sequential exchanges of resources, represented by the Incremental Centipede Game (ICG). The design comprises a "zero-end" treatment (n=164), where the ICG has a deadline yielding no payoff, and a "fair-end" treatment (n=162), where the last interaction leads to a fair outcome. The ICG, which is repeated twice, is complemented by two individual tasks: the Social Value Orientation task (SVO), to elicit social preferences, and the Cognitive Reflection Test (CRT), to assess participants' deliberation. Our results show that the fair outcome, though uncertain, prompts participants to move beyond self-interest. Indeed, an evolutionary model shows that the fair split shifts play towards a mutually beneficial strategy when players are sufficiently prosocial and use probabilistic strategies. A second game repetition reinforces this pattern, with adaptation towards cooperation. Although CRT scores do not predict behaviour in the ICG, participants who deviate from game-theoretic rationality exhibit stronger prosociality in the SVO. Notably, in the fair-end treatment, some non-rational players are individualists, suggesting that the fair split may function as a nudge. In conclusion, our findings suggest that certain personal traits, such as social preferences, allow individuals to overcome self-interest to better align with their peers and achieve better collective outcomes in mixed-motive games.

Many school choice mechanisms give students higher priority at schools they rank highly and aim to reward students' eagerness. This paper studies whether a centralized mechanism can incorporate such rank-dependent priorities while preserving truthful reporting. I consider a mechanism that jointly specifies how reported preferences modify schools' original priorities and how assignments are determined based on the modified priorities. I require the modification rule to satisfy rank monotonicity and the resulting matching to be stable with respect to the modified priorities. These two requirements jointly capture the policy objective of rewarding eagerness. My main result shows that any mechanism satisfying these requirements as well as strategy-proofness is outcome equivalent to the student-proposing Deferred Acceptance mechanism using the original priorities. Thus, if eagerness affects the outcome while the rank-dependent policy goal is maintained, the mechanism must be manipulable.

We prove two new cases of the Stuck–Zimmer conjecture. First, we settle the Stuck–Zimmer conjecture for ergodic probability preserving actions of irreducible lattices (both uniform and non-uniform) in connected semisimple real Lie groups with finite center, no compact factor and real rank at least two. We furthermore extend the Stuck–Zimmer theorem to all irreducible probability preserving actions of such groups whenever one simple factor is locally isomorphic to $\mathrm{SL}_2(\mathbb R)$. In particular, this settles the Stuck–Zimmer conjecture for irreducible actions of $\mathrm{SL}_2(\mathbb{R}) \times \mathrm{SL}_2(\mathbb{R})$. The proof relies on a new stabilizer rigidity criterion using local spectral gap of projected stabilizers in a single simple factor. Here local spectral gap is understood in the sense of Boutonnet, Ioana and Salehi Golsefidy.

Earth's lower mantle hosts large low shear-velocity provinces (LLSVPs) beneath Africa and the Pacific, but their origin remains debated. Davemaoite, a major lower-mantle mineral, has been proposed as a contributor to these anomalies; however, its experimentally measured shear modulus is approximately 30% lower than first-principles predictions, a longstanding discrepancy between theory and experiment. Here, using million-atom machine-learning molecular dynamics simulations, we show that nanoscale grain-boundary disorder, invisible to conventional x-ray diffraction, can reconcile this gap. Introducing 5.8 vol% disordered regions reduces the shear and bulk moduli by 37% and 12%, respectively. This shear-selective softening can reproduce LLSVP-like seismic anomalies in basalt-rich assemblages, but only at disorder levels corresponding to grain sizes smaller than those expected in the lower mantle. Combining the disorder-elasticity relationship with seismic constraints, we derive a minimum davemaoite grain size of approximately 100 nm. This limit is consistent with mantle grain growth models and with attenuation-based evidence for coarse-grained LLSVPs.

Data centers are increasingly exploring small modular nuclear reactors (SMRs) as a carbon-free power source, but variable datacenter demand and negative grid prices require the SMR plant to load-follow rather than run at constant output. Load following is uniquely challenging and complex for SMR plants due to underlying nuclear physics. We propose FissionReady, a datacenter scheduler that tracks each module's fuel age, xenon state, and remaining flexibility to determine per-module operating envelopes, then coordinates load-following, batch deferral, and grid purchasing through a two-timescale optimization. FissionReady achieves zero involuntary shutdowns and zero batch deadline misses while reducing water consumption by roughly one-third and grid cost by roughly half compared to a same-sized base configuration.

Electroencephalography (EEG) emotion recognition serves as a pivotal interface for human-computer interaction, yet the physiological variability across individuals complicates the already challenging task of fusing heterogeneous physiological signals (e.g., EEG and eye movements). However, most prevalent domain adaptation paradigms are tailored for unimodal scenarios, failing to address the heterogeneity of multimodal signals. Furthermore, they predominantly rely on feature-level alignment, overlooking the fundamental data-level discrepancy, which risks compromising fine-grained discriminative information during aggressive adaptation. To bridge these coupled gaps, we propose Instance-based Domain Expansion and Adversarial Learning (IDEAL), a unified framework that synergizes instance-level curriculum expansion with feature-level hierarchical adversarial alignment. IDEAL first introduces a multi-model collaborative screening mechanism, which propagates high-confidence target samples to explicitly bridge the distributional gap at the data level via a quantity-quality equilibrium strategy. We provide a theoretical analysis that this instance expansion strategy strictly tightens the upper bound of the target risk. Subsequently, a hierarchical adversarial network, augmented with the angular-contrastive constraints, progressively aligns representations from low-level statistics to high-level semantics while preserving class separability. Extensive experiments on four benchmark datasets demonstrate that IDEAL significantly outperforms state-of-the-art methods. To facilitate reproducibility and future research, our source code is publicly available at https://github.com/WY-BCI-Club/IDEAL.

Large language models are strong general-purpose code generators, but executable algorithmic trading remains a demanding specialization target: a model must translate a natural-language strategy specification into correct program logic for a specialized trading framework, execute on historical data, produce trades, and remain semantically faithful to the request. We study two complementary mechanisms for specializing language models for this setting: continued pretraining on algorithmic-trading framework code and supervised fine-tuning (SFT) on agent-validated request-to-code pairs. Evaluation is centered on QuantCode-Bench, our 400-task benchmark for Backtrader strategy generation, together with a repository-level SWE-bench-like track. Continued pretraining improves single-turn Judge Pass from 41.5% to 47.5% for Qwen3.5-397B-A17B and from 27.8% to 33.0% for Qwen3.6-35B-A3B. SFT applied after continued pretraining yields a larger gain for Qwen3.6-35B-A3B, reaching 58.2% Judge Pass and 83.5% successful backtests; in agentic evaluation it raises first-turn success from 22.3% to 58.3% and final success after up to 10 turns from 47.5% to 79.5%. Continued pretraining alone improves first-turn agentic success but lowers final success after repair from 47.5% to 32.5%, consistent with degraded instruction following, whereas SFT improves both. We also identify a capability-retention failure: domain specialization degrades parser-conformant structured tool calling, and targeted recovery SFT restores tool-call formatting but not the base checkpoint's repository-level agent performance. The results show that framework-oriented pretraining, validated SFT, and explicit capability-retention evaluation address distinct failure modes in domain-specific executable code generation.

Planktonic communities exhibit ubiquitous population distributions and patchy spatial structures, yet the fundamental mechanisms driving them remain debated. Here, we derive these regularities from a minimalistic theoretical description that incorporates stochastic fluctuations in growth rates and effective ocean dispersal. We combine global metabarcoding, microscopy, and high-resolution chlorophyll datasets and show that the decay of spatial correlations, the crossover regimes of Taylor's law, the patterns of local species diversity and biomass distributions agree with common underlying dynamics. These results suggest that the intertwined effect of diffusivity and fluctuating growth rate, captured by an emergent correlation length, shapes plankton spatial heterogeneity from local to long-range scales, reconciling local variability with macroecological patterns.