Papers

Small and medium-sized enterprises (SMEs) face increasing pressure to navigate digital transformation, yet many struggle to convert digital initiatives into sustainable organisational capability development and growth outcomes. This paper develops the Exponential SME Framework, an integrated conceptual framework that explains how SMEs may progress from digital capability development to non-linear growth potential. Drawing on literature relating to Digital Dynamic Capabilities (DDC), digital transformation practices, organisational transformation, Exponential Organization (ExO) attributes, and SME growth, the framework synthesises previously fragmented streams of research into a coherent conceptual structure. The framework proposes that Digital Dynamic Capabilities provide the foundation for Digital Transformation Practices (DASAT), which contribute to transformations across key organisational domains, including business models, value chains, customer relationships, and organisational culture. These transformations may support the adoption of ExO-like attributes, specifically experimentation, dashboards, autonomy, and community engagement, that enhance organisational adaptability, learning, and responsiveness. Through the interaction of these organisational elements, SMEs may strengthen the conditions associated with non-linear growth potential. The paper contributes to the SME and digital transformation literature by integrating multiple streams of research into a unified framework, clarifying the relationships between capabilities, practices, organisational transformation, and growth-enabling mechanisms, and providing a foundation for future empirical research. The paper also offers managerial and policy implications for strengthening SME digital transformation and growth readiness.

Large language models can produce economic policy responses that sound reasonable, but this does not show that their actions are consistent with economic dynamics. We test this by placing an instruction-tuned language model inside six Snowdrop-backed dynamic stochastic general equilibrium (DSGE) simulators. At each turn, the model observes the economy and a change in economic discourse, selects a bounded policy action, and receives the next simulated state and an economic reward. We implement a common Python interface for repeated rollouts, persistent shocks, state cloning, and rolling-horizon simulation. This setting creates a long-horizon credit-assignment problem. Policy effects may appear several quarters after an action is taken. PPO has a learned value function that can propagate delayed reward to earlier tokens through generalized advantage estimation. GRPO has no learned value function and instead assigns a group-relative advantage from complete rollout returns. It therefore cannot distinguish which earlier turn caused the outcome; if every rollout receives the same return, the normalized advantage is zero. We use PPO as the primary method and GRPO as a matched critic-free baseline. The experiments also test directional semantic signals, reward horizon, trajectory warm starts, cross-simulator transfer, and historically anchored pandemic and monetary-policy shocks. The objective is to judge policy actions by their simulated economic consequences rather than by plausible language alone.

Mainstream large language models rely on a tokenizer to encode text into a token sequence. Different tokenizers may yield token sequences of substantially different lengths for the same text. With a fixed model architecture, shorter token sequences correspond to lower inference time. We propose a tokenizer training approach named Counting and Filtering (CNF) and a text encoding algorithm called Min-Cost Encoding (MCE). MCE defines a cost function over a text segment, and determines the best segmentation by globally minimizing the overall segmentation cost. CNF builds a raw vocabulary by directly counting valid substrings, and then constructs the final vocabulary through a filtering step based on actual token usage when segmenting the training corpus with MCE. The CNF-MCE conbination offers several advantages over BPE, including higher token efficiency, greater scalability, and lower dependency. Across six text categories and two vocabulary-size groups, CNF-MCE consistently achieves better compression than the evaluated BPE tokenizers. With a 250K vocabulary, CNF-MCE increases compression rate by 26% and 30% on English web text over the o200k_base and qwen250k tokenizers. Experiments scaling the vocabulary to 1M entries on English web text demonstrate sustained improvements over BPE, with a token efficiency improvement of over 60% and vocabulary utilization rising from 52.9% to 96.9%. The MCE algorithm does not depend on a merge list (as in BPE) or token probability (as in UnigramLM), making it applicable to a wide range of vocabularies, including those built from BPE, UnigramLM, CNF, and others. Language models trained from scratch at the 1.8B and 8B scales achieve comparable average performance to models using the BPE tokenizers across 11 benchmarks. These results demonstrate that CNF-MCE can improve token efficiency significantly while maintaining competitive downstream performance.

Federated learning (FL) is a communication-efficient distributed learning paradigm. However, client drift remains one of the most critical challenges, hindering the efficient training of a global model. In this study, we propose a novel latent information sharing scheme that directly mitigates data heterogeneity across clients. Our theoretical and empirical results show that sharing a small amount of hidden-layer activations significantly improves training efficiency while preserving convergence guarantees and data privacy. Furthermore, we compare our method with existing FL approaches designed to address client drift, including FedProx, SCAFFOLD, FedPVR, FedProto, and SplitFed, and demonstrate superior model accuracy under a fixed round budget without incurring excessive communication overhead. Overall, this work presents a promising new knowledge aggregation scheme and provides a comprehensive analysis of the impact of activation sharing on federated optimization.

We develop Multivariate Quantum Signal Processing for designing matrix-valued transformations of multiple noncommuting nonnormal block-encoded matrices. Our framework equips quantum transducers with the analytic structure of multiport Infinite-Impulse-Response (IIR) filters and establishes a transfer-function methodology for designing quantum query algorithms. Efficiently computable analytic bounds that hold uniformly across the oracle promise yield efficient quantum circuits with logarithmic error dependence and near-unity success probability, paralleling the approximation-theoretic guarantees of univariate QSP/QSVT. We provide a complete characterization of achievable transfer functions and a principled library of analytic IIR components. Our modular designs greatly improve or resolve the per-oracle query complexity of open multi-oracle problems: (1) Hamiltonian simulation under $H=\sum_j H_j$ with weighted quasinorm cost $\approx t\langle\vec\lambda,\vec C\rangle_{1/2}+\mathcal{O}(\|\vec C\|_1\log(1/\epsilon))$ for block-encoding normalizations $\vec\lambda$ and costs $\vec C$; (2) sparse simulation with optimal time–norm $\Theta(\sqrt{d}\|H\|_{\rm 1\to2}t)$ and additive error dependence; (3) solving linear systems and (4) differential equations under minimal matrix assumptions with optimal weighted state-preparation and multi-term cost; (5) ground-state preparation and energy estimation; (6) generalized eigenvalue problems. We apply these to exactly simulate downfolded Hamiltonians with leading cost $t\langle\vec\lambda_{\rm eff},\vec C_{\rm eff}\rangle_{1/2}$ further reduced by fractional dynamic correlation energy contributions from coarse-grained Hamiltonians, and show similar classical-data-input-reduction results for multiscale graph classification via Kron reduction. We also show the online generalization with anisotropic Kalman filtering of streaming quantum data.

Electroencephalography (EEG) foundation models are often pretrained with a fixed channel vocabulary or a limited set of montages, making transfer difficult when electrode layouts change. We propose CortexBridge, a lightweight adapter that combines EEG features with electrode and atlas coordinates to map arbitrary montages into a shared cortical latent space. Evaluated with three frozen foundation models on five brain-computer interface (BCI) datasets from the Mother of All BCI Benchmarks (MOABB), CortexBridge improves performance in 13 of 15 evaluations. The gains in balanced accuracy average 0.80% for EEGPT, 0.70% for LaBraM, and 3.26% for CBraMod, with a maximum gain of 13.02% on 12-class steady-state visual evoked potential (SSVEP) classification. Visualizations of the learned atlas representations reveal task-dependent spatial patterns, with SSVEP showing a more concentrated representation in the Yeo Visual network than auditory P300. These results establish cortical alignment as a learnable and anatomically grounded routing mechanism from heterogeneous EEG montages to pretrained foundation models.

The 35-day superorbital modulation of the X-ray binary Her X-1 was systematically investigated using a 35-year multi-instrument X-ray observational baseline. We developed a Rank-Searching method using the peak intensity of the main high state to identify anomalous low states (ALSs) with an objective and reproducible criterion. The method recovers the known ALSs, identifies two previously unrecognized candidate episodes, ALS 1991 and ALS 2015, and independently recovers the recently reported ALS 2025. We then characterized the long-term superorbital phase evolution, identifying seven epochs that alternate between smooth and variable behavior. The smooth component is empirically described by a cubic trend, while the recurrent variable epochs have a characteristic timescale of approximately 3400 days. Within a first-order tilted-disk framework, the secular phase evolution corresponds to an inferred cumulative disk-mass increase of $\sim1.8\%$ and an outer-disk-radius expansion of $\sim0.9\%$, while the short-term phase excursions can be reproduced by disk-mass variations of $\lesssim2.4\%$. A viscous diffusion timescale of approximately 168 days provides a characteristic timescale for the short-term disk response. Finally, the stable phase relation and high profile similarity between the main and short high states indicate that the warped disk maintains a remarkably stable global geometry over the observational baseline.

We equip Khovanov's colored $\mathfrak{gl}(2)$-homology groups with a degree-2 derivation. Over a field of characteristic $p>2$, the derivation specializes to a $p$-nilpotent differential, which leads to a categorification of colored $\mathfrak{gl}(2)$-homology at a $2p$th root of unity.

Gibbs samplers, and more generally component-wise Markov chain Monte Carlo algorithms such as Metropolis-within-Gibbs, can be implemented using either random-scan or deterministic-scan updates. How much convergence can depend on this choice of scanning rule has been a longstanding question. We study this problem through $L^2$ spectral gaps, measuring computational cost in units of component updates. For a $d$-block Gibbs sampler, the cost of random scan is at most twice that of any fixed deterministic scan, while the reverse cost ratio is at most of order $d^2$. We extend these comparisons to general reversible component-wise updates under the global block-wise contraction condition. If $K_j$ denotes the update of block $j$ and $P_j$ its Gibbs counterpart, and $\|K_j-P_j\|\leq \lambda_0<1$ for $j=1,\ldots,d$, then the cost of random scan is at most $2/(1-\lambda_0)$ times that of deterministic scan, while the reverse cost ratio is of order at most $d^2/(1-\lambda_0)$. Examples show that the cost bounds for random scan relative to deterministic scan are asymptotically sharp and that the joint dependence on $d$ and $(1-\lambda_0)^{-1}$ in the reverse comparison cannot be improved uniformly. This work was assisted by generative AI, including for formal verification of mathematical results in Lean. The human author reviewed and verified the mathematical content and takes full responsibility for the results.

We propose an antithetic coupling that reduces the computational cost of gradient random-walk methods for one-dimensional convection-diffusion equations. The methods represent the solution's spatial derivative by particles with positive or negative masses, whose cumulative sums reconstruct the solution, and model diffusion by Gaussian displacements. The particles reorder and, for a nonlinear flux, interact. Antithetic sampling negates the displacements of a second simulation, which positively correlates the contributions of particles matched by position when their masses have opposite signs. Our coupling instead gives matched particles of opposite sign identical displacements and retains opposite displacements for equal-sign pairs. We prove that each simulation keeps the distribution of an uncoupled run as the matching changes, so the averaged solution keeps the bias of one simulation. For fixed particle configurations and matching, the coupling minimizes each matched pair's conditional variance contribution after diffusion among all couplings with the same Gaussian marginal distributions. From the same configurations, its variance reduction over antithetic sampling at the final diffusion stage has a closed form for any flux. For the heat equation, a martingale decomposition extends both results to each step's contribution to the terminal variance. In a Burgers' equation test, the coupling meets a tolerance for the spatially integrated mean squared error with 4 simulations, against 34 for antithetic sampling and 65 for independent sampling. In a second test, it meets a common error tolerance in about 40% less time than pairing particles only within each sign.

Designing high-performance microwave absorbers requires specialized expertise in electromagnetic theory, materials science and simulation programming, and entails time-consuming optimization. Here, we present AbsorbEvo, an agentic framework for autonomous inverse design that translates natural-language performance objectives into designs verified by full-wave simulations. Its candidate evolution strategy integrates language reasoning, physics-based prediction and historical feedback. A large language model proposes the directions and magnitudes of parameter adjustments based on task objectives and computational history. The system combines directed increments with global sampling to generate candidates and uses a low-cost predictive model as a physics prior to rank them. Only high-ranking designs undergo full-wave simulation. Results passing physical validity checks are used to evaluate performance and guide subsequent search. Experience from training tasks is further distilled into textual skills, which are independently validated before use in new tasks. Under identical proposal budgets on held-out AbsorbBench-36 tasks, AbsorbEvo achieved a task success rate of 79.17%, versus 25.00% for a generic agent and 12.50% for random search. Its mean best coverage was 0.7816, compared with 0.6434 and 0.6448, respectively. By integrating language reasoning and physics-based feedback into design decisions, AbsorbEvo provides a methodological foundation for natural-language-driven autonomous inverse design of microwave absorbers.

A long-term conversational assistant must recall the right memory at the right moment, yet the memory that matters most is often not similar to what the user says now. Current systems recover such associations by letting an LLM reason at write or read time, at a cost of hundreds to over a thousand LLM calls per memory bank and up to several thousand context tokens per query. We argue that association is a learnable relevance: the pointwise mutual information of memories under how human lives unfold. We introduce Madeleine, which learns amortized association: offline, an LLM life simulator writes simulated lives, whose cue-trigger pairs teach a query encoder a residual association on top of frozen similarity; online, it calls no LLM and plugs into any vector memory by replacing only the query encoder. On LoCoMo-Plus under the official protocol, Madeleine (I) reaches 66.6 when plugged into HyperMem, the highest among all systems evaluated under this protocol; (II) used alone, reaches the score of HyperMem as released (52.4 vs. 52.9) with zero LLM calls and about 1/21 of its answer context; and (III) lifts T-Mem by 26.2 points, significantly outperforms the same untrained backbone inside both systems, and leaves ordinary QA intact on the 4B backbone.

In neutron-star and dense quantum chromodynamics (QCD) physics, it is important to understand how baryonic matter evolves into quark matter at high density. A continuous hadron–quark crossover is one of the most attractive candidates because it can reconcile the comparatively soft equation of state around nuclear density with the rapid stiffening required to support massive neutron stars. Recent neutron-star observations motivate equations of state that stiffen rapidly across the intermediate-density regime, while quarkyonic-matter descriptions feature a shell-like baryon momentum distribution there. In this paper, we review our recent work on the microscopic description and interpretation of these phenomena by drawing an analogy with the Bose–Einstein condensate (BEC) to Bardeen–Cooper–Schrieffer (BCS) crossover in ultracold Fermi gases and two-color QCD. As pairing fluctuations play a crucial role in the BEC–BCS crossover, we discuss the role of its three-body counterpart, that is, tripling fluctuations associated with baryon formation in the hadron–quark crossover. In terms of a phase-shift representation of tripling fluctuations, the interplay between a three-body bound-state pole and the scattering continuum suppresses low-momentum baryonic occupation, generates a baryonic momentum shell, and reduces the density susceptibility, thereby producing a peak in the speed of sound. We review the demonstration of this mechanism in a one-dimensional three-component Fermi gas, its relativistic extension, and its connection to phenomenological quarkyonic equations of state.

As the capabilities and ubiquity of Large Language Models (LLMs) grow, so does their environmental footprint. Despite calls for responsible AI, the machine learning community lacks standardised practices for carbon accounting. Our automated literature review of the 5,285 papers accepted to NeurIPS 2025 reveals that reporting of environmental impact is nearly non-existent. To catalyse a shift toward sustainable AI, we define standardised sustainability metrics for evaluating model training efficiency, accompanied by simple heuristics to estimate the carbon cost of LLM inference. We implement these metrics in carbonbenchmark, a drop-in software solution for tracking and reporting emissions. Finally, to combat the pursuit of marginal accuracy gains at disproportionate environmental costs, we formalise the `Smallest Model that Achieves the Job' (SMAJ), a framework which challenges the field to prioritise computational efficiency and environmental accountability alongside traditional `State-of-the-Art' (SotA) accuracy.

Statistical validation takes time, and alpha can decay before the evidence justifies deployment. We quantify the surviving opportunity through Certified Alpha Capacity: the maximum expected value remaining under prescribed false-deployment and power constraints. In a canonical Gaussian experiment, we derive an exact certification frontier and the sharp law $\mathcal K_{\alpha,\beta}/\log(1/\varepsilon)$ for residual information capacity just above it. Near the frontier, a single interim decision can preserve orders of magnitude more capacity than fixed-time testing. Mapping information into cash reveals why equally certifiable signals can retain different economic values. In the specified competitive equilibrium, lifetime information approaches the certification frontier exponentially as research costs vanish.

Gaussian splatting has recently emerged as an efficient representation for images and videos due to its explicit structure and fast rendering capability. Existing Gaussian-based video representations often decompose a video into canonical Gaussians and temporal deformation. However, when a video contains large global motion such as camera movement, the canonical representation may become misaligned with individual frames, increasing the burden on the temporal deformation model. In this paper, we propose an affine-atlas canonical Gaussian representation, which constructs canonical Gaussians in a larger affine-aligned atlas space. Frame-wise affine transforms absorb global motion before canonical Gaussian construction, reducing the gap between the canonical representation and target frames. Since the proposed method only modifies the canonical construction stage, it can be integrated into existing canonical-Gaussian-based methods with negligible additional parameter cost. Experiments show that our method improves reconstruction quality especially for sequences with large camera motion.

We present a novel numerical integration method for non-periodic functions over the high-dimensional unit cube $[0,1]^s$ by a specially crafted (unequally-)weighted quadrature rule using transformed lattice points, with the additional option of subsampling. The method is designed for integrands whose mixed derivatives up to order $\alpha$ are square-integrable with respect to a product Chebyshev density. The ingredients are (i) periodizing the integrand, (ii) approximating the smooth periodic function by a kernel interpolant at lattice points, (iii) integrating exactly the product of the kernel interpolant and the nonsmooth density, together with subsampling alternatives for (ii) and~(iii). With $|J|$ corresponding to the number of function evaluations, we achieve the convergence rate close to the order $|J|^{-(\alpha-1/2)}$ and $|J|^{-\alpha}\,($1.110721$)^s$, where the implied constants are independent of $s$ under favorable conditions. We can interpolate these two results to trade between the convergence rate and the growth in $s$. We provide numerical experiments to demonstrate our theory.

We discover a universal trade-off between the dissipation required to maintain a quantum state more sensitive than a thermal one and the sensitivity gain for a driven single-mode bosonic system under a thermal environment. Specifically, for Gaussian states, we rigorously derive a lower bound on the entropy production rate in the steady state in terms of the quantum Fisher information gain relative to a thermal reference state. Numerical investigations of Kerr nonlinearities further support that these bounds can hold beyond the Gaussian regime. Our results establish a new connection between steady-state thermodynamics and quantum metrological performance.

We propose and study a compartmental epidemiological model that incorporates both waning immunity and mutation-driven immune escape into a single-strain framework. The model, termed SIRCm, extends the SIRC (Susceptible-Infected-Recovered-Cross-immune) model, in which recovered individuals pass through an intermediate cross-immune class before returning to full susceptibility. In SIRCm, both the rate of immune waning and the rate of immune escape are amplified by a feedback function tied to the current level of infection. We study two formulations of this feedback, a transmission-driven model, in which mutation opportunities arise at the point of new infection, and a prevalence-driven model, in which they arise in proportion to the infected population alone. For both models and any feedback strength, we prove existence of an endemic equilibrium using persistence theory, and characterize its behavior analytically in the weak and strong-feedback limits, recovering the SIRC and SIS models, respectively, as limiting cases. Unlike the SIRC model, we show that the SIRCm endemic equilibrium is unstable in certain parameter regimes. This leads to novel behavior, even in the absence of seasonal forcing.

How much can test-time computation improve a language model, and at what cost? Test-time scaling is widely proposed as a substitute for larger models, but existing comparisons mostly evaluate one domain at a time and rarely charge selection to the budget. We introduce SELF-POT, a benchmark and evaluation framework that measures the test-time potential of a model across competition mathematics, competitive programming, and agentic workflows. SELF-POT separates candidate coverage from final accuracy on static tasks, tracks correctness transitions under revision, and measures protocol completion alongside task success in agentic environments. Under a unified budget rule, it compares Direct inference with parallel sampling and self-revision under fixed multiples of the Direct budget, and charges every model call, including selection and critique, in dollars. This design supports two kinds of comparison: the gain a model obtains from additional inference, and a lower-cost model with additional inference against a stronger model. Across five low-cost reasoning models on 350 sealed tasks, with Claude Opus 5.5 Direct as the reference, the returns depend on the domain, the selection rule, and failure handling. When we replay the retained programming candidate pools, public-example selection raises correct submissions from 376 to 453 of 500 scheduled cells while saving 12-49% of logical API cost across models, and simply retaining an available candidate when judging fails recovers 61 submissions at unchanged cost. On identical mathematics pools, judging with fallback yields 186 correct submissions versus 182 for voting, while voting saves 12-21% of logical API cost. These controlled replays show how selection and failure handling change the gains realized from the same generated candidates, and they quantify the marginal value of a model judge.

We study the emergence of cuspons through wave breaking, with the Camassa–Holm equation as a principal model. For a broad open class of smooth initial data, the first singularity occurs at a unique point and has the sharp local Hölder regularity \(C^{3/5}\). We construct a continuation of the Lagrangian flow through this singular time and reconstruct from it a conservative Eulerian weak solution. We show that the first singularity does not simply persist after breaking; instead, it bifurcates immediately into an anticuspon–cuspon pair, each having the sharp local Hölder regularity \(C^{2/3}\), with opposite one-sided slope orientations. The two singular branches move apart from the original breaking point, and the portion of the profile between them reverses its local monotonicity. The proof is based on an extended inverse-slope variable that remains regular through wave breaking. At the first singular time, this variable has a nondegenerate quadratic zero, which splits immediately afterward into two simple zeros. The corresponding degeneracy of the Lagrangian flow map converts the quadratic first-breaking geometry into the \(C^{3/5}\) profile and the post-breaking simple-zero geometry into the \(C^{2/3}\) cuspon profiles. The generalized Hunter–Saxton family provides an explicit realization of the same Lagrangian mechanism and also allows us to track the subsequent interaction and annihilation of the singular branches. Thus an anticuspon–cuspon pair emerges dynamically from smooth initial data through wave breaking, rather than being imposed through a singular traveling-wave ansatz.

Retrieval-based speculative decoding (SD) drafts tokens by copying continuations from existing text, which suits coding agents that repeatedly reproduce code, logs, and earlier attempts. Yet existing methods fall short in agent pipelines: much of the reusable text is missing from their corpora or stored in a form that differs from what the agent emits, and their draft lengths ignore that accept length varies across agents and drifts over turns. We present AgSpec, a framework that supplies the corpus and draft-length policies that existing retrieval engines lack in coding-agent pipelines. AgSpec retrieves from session, workspace, and global corpora, retaining the ongoing session trajectory and indexing opened files in the agent's emission format. It bounds each agent's draft length with an offline-profiled cap and adapts the length online from verification feedback. On two repository-level multi-agent coding benchmarks, AgSpec outperforms five retrieval-based drafters and EAGLE-3 in most evaluated settings, raising generation throughput over autoregressive decoding up to 4.37$\times$ at batch size 1 and 4.76$\times$ at batch size 16. AgSpec also remains effective on benchmarks without a repository or a multi-agent pipeline, showing that its gains generalize to coding agents broadly.

This letter presents a method to compute fundamental bounds on the total efficiency of pixel antennas. Using semidefinite relaxation, we obtain an upper bound on the optimization problem of maximizing the antenna's total efficiency subject to binary-valued pixel states. The bounds are demonstrated for antennas with varying numbers of pixels and compared with realizable solutions obtained using the genetic algorithm.

Asset names are associated with investor trading and asset prices, but where names and issuer quality are formed jointly, information, preference and attention-coordination explanations are difficult to distinguish. We study a token launchpad on which tokens issued under the default template use the same contract code, have a fixed supply and carry no cash flows, while names can be registered at almost no cost, cannot be verified and may be reused. Using all 458,174 default-template tokens launched from February to June 2026 and a sample split by creator entity fixed in advance, we find that the share of tokens attracting an outside buyer rises from 34.9% to 52.8% across name-appeal deciles, but within creator and creation minute the effect of appeal is small and does not replicate in the validation sample. Conditional on early capital inflow and creator fixed effects, the residual effects of names on migration, peak market capitalization and post-peak drawdown are statistically equivalent to zero, and buyers who follow appealing names earn no economically meaningful premium. The same name attracts less capital with each reuse, names whose first token draws more capital are copied sooner, and competition over names mainly transfers wealth among buyers.

Autonomous robots on one-shot missions run over horizons far longer than the trajectories seen during training, under a fixed onboard compute budget. We present FRANK, a 507K-parameter recurrent architecture that combines tau-gated recurrent modules, content-addressable memory, and a feedforward reflex pathway. We evaluate it against recurrent, state-space, and reduced modular baselines at matched parameter count on four algorithmic sequence tasks, trained at length 5-20 and evaluated out to two million tokens. At 100,000x the maximum training length, 6 of 10 FRANK seeds retain exactly 100.0% accuracy, while none of the 50 baseline configurations does, five architectures at ten seeds each with none left incomplete (Fisher exact, two-sided p = 4.2E-6. Targeted lesions across the four tasks yield four distinct component-reliance profiles, consistent with task-dependent allocation across the recurrent, memory, and reflex pathways. Separately, a FRANK policy trained in simulation drives a physical ground vehicle to commanded waypoints through obstacles without teleoperation.

In this study, we construct a bayesian holographic QCD model by integrating the Einstein-Maxwell-Dilaton (EMD) framework with lattice quantum chromodynamics (QCD) data at zero chemical potential, specifically entropy, the square of the speed of sound, and baryon number susceptibilities, while systematically incorporating error estimates from the lattice QCD (LQCD) results. Leveraging a bayesian inference framework, we first achieve a precise calibration of the model parameters, then perform a comprehensive investigation into the thermodynamic properties of \((2+1)\)-flavor QCD at both zero and finite chemical potentials, and finally provide a prediction for the location of the critical end point (CEP) in the QCD phase diagram. Our results indicate that, under the maximum a posteriori (MAP) estimation, the CEP is located at $(T, \mu)_{\mathrm{MAP}} = (0.119, 0.615)\mathrm{GeV}$. Furthermore, we provide the predicted regions for the CEP at $68\%$ and $95\%$ confidence levels (CL), yielding $(T, \mu)_{68\% \mathrm{CL}} = (0.1176\text{--}0.1204,0.59\text{--}0.63)\mathrm{GeV}$ and $(T, \mu)_{95\% \mathrm{CL}} = (0.1172\text{--}0.1206,0.58\text{--}0.63)\mathrm{GeV}$, respectively. A thorough comparison with predictions from other theoretical models validates the robustness and predictive power of our approach. This work not only establishes a novel analytical framework for holographic modeling but also provides valuable theoretical insights into the phase transitions of strongly interacting matter under extreme conditions.

Vocal behavior provides important signals for speech-based depression detection, but its interpretation often depends on what is being said and how it functions in context. However, existing methods typically treat acoustic cues as context-independent markers, making it difficult to distinguish vocal form from its context-dependent communicative function. We propose ParaCalib, a framework that semantically calibrates paralinguistic behavior by interpreting vocal patterns relative to utterance-level semantic context and inferred communicative function and representing them in a comparable state space. Concretely, ParaCalib uses an Audio-Language Model (ALM) to generate contextualized vocal descriptions and an LLM-based Paralinguistic State Extractor (PSE) to map these descriptions into a structured Semantically Calibrated Paralinguistic (SC-Para) representation. ParaCalib achieves the highest mean Macro-F1 among the evaluated methods, reaching 71.9% on DAIC-WOZ and 90.5% on MODMA. Our controlled analysis provides direct evidence of semantic calibration at the PSE stage: under a fixed caption-level acoustic description, varying the accompanying semantic context changes the inferred depression-related paralinguistic evidence. Our exploratory analysis further identifies recurring configurations of paralinguistic states associated with depression labels rather than a single uniformly dominant state.

Coordinated multi-humanoid loco-manipulation is promising yet challenging due to high-dimensional whole-body control, decentralized decision making, and scalability. While recent reinforcement learning methods have improved single-humanoid whole-body control, extending them to the multi-humanoid setting remains nontrivial and often requires substantial reward engineering or task-specific design. We propose MASkillBlender, a general multi-agent reinforcement learning framework to achieve decentralized multi-humanoid whole-body coordination. By learning a shared decentralized high-level policy over reusable pre-trained single-humanoid skills, MASkillBlender enables coordinated behaviors using only task-level rewards, without requiring task-specific motion references. To improve learning efficiency, we further introduce a permutation-based data augmentation strategy for homogeneous multi-humanoid systems, and theoretically show that the permuted samples preserve the policy-gradient direction of the original samples under the homogeneous Markov game formulation. We evaluate MASkillBlender on multiple multi-humanoid coordination tasks across two humanoid embodiments. Simulation results demonstrate that the proposed framework consistently achieves strong task performance and enables coordinated behaviors across different tasks and humanoid embodiments.

Dual resonance reconstructs a scattering amplitude from pole expansions in different channels. We propose analytic dual resonance (ADR), requiring that the complete leading pole sum at an integer resonance admit an absolutely convergent local contour integral. ADR yields one-dimensional pairing relations between residues in arbitrary triangulations. Together with an integer planar spectrum, a joint crossing-spin bound, and complex Regge decay permitting unsubtracted reconstruction, these relations uniquely fix the scalar string amplitude for every $n\ge5$. ADR also predicts higher-dimensional pinch relations, which we examine at five and six points. These results connect the analytic structure of dual resonance to the uniqueness of string interactions.

Spin qubits are an attractive platform for scalable quantum computation, and recently, laterally gated quantum dots (QDs) in Ge heterostructures have emerged as a leading implementation. Radio-frequency (RF) single-hole transistor (SHT) spin readout and integrated superconducting resonator charge readout have demonstrated in Ge QDs, but their wiring overhead and on-chip size can pose challenges, while gate dispersive (GD) readout, which requires less on-chip overhead, has yet to be demonstrated for Ge QDs. Here we demonstrate compact, low-overhead RF-GD charge readout of a Ge double QD (DQD), both directly, and assisted by a proximal single-hole box (SHB). We resolve QD loading and interdot transitions with both techniques. We obtain unity signal-to-noise ratio (SNR) in 2.7 millisecond and 135 microsecond integration time, for gate-dispersive detection of the DQD and SHB-DQD systems, respectively. We find that the SHB enhances the interdot signal-to-noise ratio (SNR) of our 540 MHz circuit by a factor of 20 for a DQD coupling $t_c \approx 5$ GHz. Power and temperature dependent measurements verify that SNR is limited by dielectric loss. We argue that with achievable improvements in dielectric loss and amplification, high-fidelity (99%) readout in less than 1 microsecond should be possible, assuming ideal spin-to-charge conversion, which has seen rapid recent progress. These results establish a simple and compact gate-based readout approach for direct and SHB-enhanced readout of Ge heterostructure QDs.