Papers

Renewable energy power plants are increasingly expected to provide ancillary services to the grid. Yet, the variability of sources such as solar photovoltaics requires additional operational flexibility to deliver these services. In this context, hybrid energy storage systems (HESSs) offer an appropriate solution, because different technologies with complementary characteristics can be applied and leveraged to share the power demand and improve the overall performance. Yet, a proper power sharing requires a detailed model of the storage elements. This aspect has barely been studied in the literature, where most models consider constant efficiencies. Moreover, most formulations are based on optimisation problems that are computationally too demanding to be solved in real time. In this paper, a controller is proposed to minimise the losses of a HESS considering detailed power-dependent efficiency curves of each storage technology. The power sharing method is based on the analytical verification of the Karush-Kuhn-Tucker (KKT) conditions, which makes it computationally efficient and suitable for real-world deployment. The proposed controller is applied to a test case consisting of a 10 MW PV power plant with a HESS based on a lithium-ion battery and a redox-flow battery, each with 2.5 MW power and 5 MWh capacity. The main contributions are verified via numerical simulations performed in MATLAB/Simulink, whereas a real-time implementation deployed in OPAL-RT demonstrates the viability of the algorithm for real-time applications.

We study collision-free movement problems on graphs, where the task is to coordinate a set of robots so that they reach a target formation satisfying a desired property while minimizing the total travel distance. This framework extends two classical models: (a) minimizing movement [Demaine et al., TALG '09, '14], which does not enforce collision avoidance, and (b) coordinated motion planning or multi-agent path finding [Eiben et al., SoCG '23, Deligkas et al., ICALP '24, among many others], where each robot is assigned an explicit target position. We focus on the setting where the target formation of the robots should be connected. We analyze the parameterized complexity of the problem with respect to the number of (main) robots and the total travel length on grid graphs and two natural generalizations thereof: planar graphs and unit disk graphs.

We introduce effective depth ($\Deff$), a scalar diagnostic that treats the layer-wise residual stream of a transformer as a discrete-time process, measures how representation similarity decays with layer distance, and aggregates that profile into one number. Across sixteen decoder-only language models, $\Deff$ separates a structural consequence of residual accumulation from an empirical one: even maximally diverse orthogonal updates have the closed-form reference $F_L = 2L/(L+1)<2$, yet fifteen of sixteen default measurements lie below $F_L$ (Qwen3.5: 32–44%, OLMo-2: 40–41%, Pythia: 23–28%). Matched references show that the gap is not caused by the persistent initial state or update-size imbalance, but is largely a calibrated signature of correlated residual updates rather than evidence that depth is unused. Symmetric position-0, token-normalisation, and top-PC controls show the regime is not reducible to BOS or top-PC artefacts: the lone above-reference default outlier joins the same regime, and all sixteen models are sub-reference after token-normalisation or top-1-PC removal. Intermediate checkpoints show that the regime is established early in OLMo-2 and stable through 5T tokens, while Pythia-1.4B follows a distinct decreasing trajectory. A controlled residual-carry intervention supports the mechanism, and $\Deff$ is best read as a global accumulated-state diagnostic, not as a capability score or pruning method.

We give a criterion for the existence of a global supporting affine functional for the entanglement of formation at an entangled two-qubit state. It exists if and only if the spin-flip bilinear form is nondegenerate on the support of the state, or equivalently, if the last concurrence root associated with the support is nonzero. In this case the entanglement of formation satisfies a two-sided Lipschitz bound relative to the reference state. In the degenerate case the state has rank two or three, and its projective support is tangent to the product-state quadric. By suitably choosing a perturbation outside the original support space, the decrease in the entanglement of formation is, to leading order, proportional to the square root of the mixing parameter. We give an explicit expression for the leading coefficient. Such a square-root decrease violates the one-sided Lipschitz bound, and therefore no globally supported affine functional exists at this state.

We prove that finite-dimensional real normed spaces with isometric unit spheres are linearly isometric, where each sphere carries the distance induced by the ambient norm. This answers the object-level question of Kadets and Martín in finite dimensions, without asserting linear extension of a prescribed sphere isometry. We associate with each norm a family of Plücker contraction bodies generated by volume-normalized maximal minors of linear contractions into finite-dimensional $\ell_\infty$ spaces. Null-Lagrangian identities show that the sphere metric determines these bodies. A finite exposed-face construction recovers almost-norming contractions from the convexified data, and compactness together with an exact volume identity yields a linear isometry.

AI assistance can improve performance without improving self-assessment. We report a study (N=366) comparing Human alone and Human+AI performance on reasoning tasks, for which the AI model is benchmarked on the same items. Participants estimated global and block performance and rated confidence in their answers. Human+AI achieved higher scores, but self-estimates tracked performance weakly. Average overestimation was similar across groups, covering individual errors. Across tasks, confidence distinguished correct from incorrect answers less accurately in the Human+AI group, while within-task differences remained uncertain. The Dunning-Kruger pattern was found in both groups, with a larger observed contrast in Human+AI. Controls for score noise reduced but did not eliminate the pattern, with the controlled group difference remaining inconclusive. An extended computational account describes global and block estimates. Our findings distinguish performance augmentation from metacognitive augmentation and motivate interfaces that support verification, communicate task-specific AI model performance, and help users evaluate the quality of their joint work rather than produce answers.

Quantum algebras ($q$-algebras) have been used in the literature to deform spin-chain models. In particular, in previous works by the present authors, the Kittel–Shore Hamiltonian was deformed using $\mathfrak{su}_q(2)$ algebra. The deformed Hamiltonian was presented for a general spin, and the thermodynamic properties were studied for spin-1/2 particles. In this paper, we focus on the $q$-deformation of the Kittel–Shore model for spin-1 systems. We analyse the impact of this deformation on the thermodynamic properties and phase transitions of the system as a function of the deformation parameter $q$. Specifically, we study the specific heat, magnetic susceptibility, and magnetisation, including a detailed analysis of the Curie temperature characterising the ferromagnetic phase. Furthermore, a finite-size scaling analysis is conducted for both the ferromagnetic and antiferromagnetic cases. Results are systematically compared with those of the undeformed model to reveal the physical effects of the $q$-deformation.

Scaling language models to long contexts is limited by the quadratic cost of self-attention. Block sparse attention offers an efficient alternative, but selecting the retained blocks remains a bottleneck. Conventional block selection requires scoring all query-block pairs and therefore remains quadratic in sequence length. To address this issue, we propose PISA, a block-sparse attention mechanism that employs a pyramid Top-$K$ selection strategy. The main idea is to gradually narrow down the candidates across different levels, making it more efficient to find the most relevant keys. Specifically, we construct a coarse-to-fine hierarchy of keys and perform selection from the coarsest level. At each level, LogSumExp scoring is applied to a bounded candidate set to select candidates for the next finer level, continuing until the finest level is reached. Through pooling, we construct $O(\log N)$ levels of keys, yielding an overall complexity of $O(N\log N)$, where $N$ denotes the sequence length. We develop hardware-aware Triton kernels for both training and inference, fusing hierarchical routing and LogSumExp scoring without materializing the query-key score matrix. We further evaluate our method on language modeling tasks. Compared with the baseline, our method achieves comparable performance on benchmarks such as commonsense reasoning while delivering better results on retrieval tasks.

The slowdown of Moore's law and the area limit of monolithic integration have made chiplet-based designs inevitable across many domains, including network ASICs. However, combining multiple network ASICs together poses a fundamental challenge: maintaining sufficient inter-ASIC bandwidth to match the performance of an idealistic single-ASIC design. Providing full bandwidth is prohibitively expensive as it requires valuable forwarding capacity, while reducing inter-ASIC bandwidth creates severe performance bottlenecks. We propose a novel multi-ASIC switch architecture that introduces a circuit-switched indirection layer in front of the ASICs. This layer flexibly remaps ingress ports across ASICs, localizing traffic and minimizing inter-ASIC communication based on observed patterns. Our system, Fastroute, combines packet and circuit switching to deliver performance comparable to a single-ASIC switch while reducing inter-ASIC bandwidth requirements. This frees up capacity for external network interfaces, allowing Fastroute to outperform traditional non-oversubscribed multi-ASIC designs. Our hardware prototype demonstrates the system's functional feasibility by evaluating it on an LLM training workload. By reducing bandwidth and power overhead, Fastroute bridges the gap between silicon fabrication limits and soaring application demands. It provides an efficient transition to multi-ASIC switches, enabling bandwidth and radix demand to be met without waiting for the next ASIC generation.

For linear-quadratic $N$-dimensional Nash systems with mean-field-like scaling and strongly displacement monotone data, we prove uniform-in-time a priori derivative estimates exhibiting the expected scaling as $N \to \infty$ in order to pass the system to the limit. As a consequence, in a heterogeneous (i.e., non-symmetric) mean-field setting, we obtain both uniform-in-time convergence of the Nash system on any arbitrarily long horizon $[0,T]$ (as $N \to \infty$), and convergence of both the $N$-dimensional Nash system and the corresponding master equation to their respective ergodic counterparts (as $T \to \infty$).

We study fundamental information-theoretic limits of robust stochastic optimization when the distribution is known only through its exact moment sequence. We develop a unified framework that produces families of distinct distributions sharing all moments yet inducing radically different optimal decisions, thereby establishing strong impossibility results for a range of decision problems under moment ambiguity. Our binary and $N$-way moment-matching constructions separate quantiles, threshold regions, and online selection opportunities while preserving every moment. For the newsvendor problem, moment equivalence causes quantile ambiguity, inducing any fixed or randomized order quantity to fail arbitrarily badly. For revenue maximization, no deterministic or randomized posted pricing scheme can secure a nontrivial approximation relative to the full-information benchmark. For the secretary problem with cardinal observations, the optimal robust success probability under exact moment disclosure is the classical finite-horizon value $s_m$, even for i.i.d. observations. This value converges to $1/e$, compared with the full-information limit of approximately $0.5801$ due to Gilbert and Mosteller (J. Am. Stat. Assoc., 1966). We also recover and expand upon the recent impossibility result of Correa et al. (STOC, 2026) for prophet inequalities with moment knowledge. For every fixed $r\ge1$, exact moment knowledge can yield at best a $\Theta(1/\log m)$ competitive ratio, even when competing against relaxed benchmarks based on expected $r$-th order statistics or when the algorithm is allowed to select $r$ items.

We prove the (uniform) Yau-Tian-Donaldson Conjecture for transcendental Kähler classes in the case of trivial automorphisms. Indeed, we show the existence of Special Kähler Fujita Approximations of big cohomology classes associated to big test configurations, establishing the equivalence between the uniform $K$-stability and the strengthened version for models. More generally, we show that any big cohomology class on a compact Kähler manifold admits such Special Kähler Fujita Approximations: the volume and the analogue of the Riemann-Roch coefficient of the big class are both approximated by those of Kähler classes on higher compactifications.

Pinching-antenna systems (PASS) create reconfigurable wireless channels by moving dielectric antennas along a waveguide. Mechanical position errors translate into phase errors that degrade performance, yet almost all prior PASS studies assume ideal placement. This paper models this position error, derives the resulting phase error, and analyzes its impact on time-division multiple access (TDMA) and non-orthogonal multiple access (NOMA), obtaining their ergodic-rate bounds and closed-form outage probabilities. Analyses and simulations reveal that single-antenna TDMA is immune to position error, whereas multi-antenna TDMA loses array gain and NOMA suffers an interference floor that cannot be overcome by increasing transmit power.

Cryptographic hash functions over integers modulo a prime play a decisive role in the efficiency and security of proof systems for computational integrity. Early designs focused on compact arithmetic circuits and efficient software execution, primarily targeting general-purpose CPUs rather than hardware accelerators. This work focuses on enabling efficient resource sharing and hardware acceleration alongside efficient software execution. We propose BenX, a permutation-based hash function designed for hardware acceleration, fast CPU execution, and low circuit complexity in proof systems. To construct the underlying permutation, we turn the Benes network into an invertible function over $\mathbb{F}_p^2$ using a Dickson polynomial. Its structure yields a permutation particularly suited to hardware resource sharing and acceleration. Fully pipelined on FPGA, BenX matches the throughput of Poseidon2, with 1.36x and 2.15x lower latency for Goldilocks and BabyBear, respectively. Although larger as a standalone core, it makes more effective use of shared hardware: in our dual-mode architecture, where the NTT and hash share field multipliers, BenX keeps 100% of them busy, compared with 6.25-20.3% for the partial rounds of Poseidon and Poseidon2. In software, BenX outperforms Poseidon in all tested 31- and 64-bit configurations, but is slower than Poseidon2, except for the 16-element BabyBear instance. In zero-knowledge proofs, BenX proves Goldilocks permutations 6-10x faster than Monolith. Its compact arithmetization uses about 36% fewer trace cells than Poseidon and Poseidon2 over BabyBear, while its fast variant requires 1.3-2.6x the prover time of Poseidon2.

This paper studies constrained optimization problems through the lens of feedback control. Building on the interpretation of Lagrange multipliers as feedback controllers, we propose the saddle-point PID (SPPID) dynamics: a unified proportional–integral–derivative (PID) framework for continuous-time saddle-point dynamics. The proposed dynamics employ PID control for equality constraints and anti-windup PI control for inequality constraints. We show that SPPID is equivalent to a preconditioned primal–dual gradient flow of the augmented Lagrangian. This equivalence reveals the distinct role of each feedback component: integral action enforces constraint satisfaction, proportional action induces the augmented Lagrangian structure, and derivative action modifies the geometry of the primal dynamics via a state-dependent Riemannian metric. For convex problems, we establish convergence to the KKT set together with conditions for global asymptotic stability. For equality-constrained problems, we establish local exponential convergence for nonlinear equality constraints under standard assumptions, and show that projected gradient flow emerges as the infinite derivative-gain limit. For strongly convex problems with affine equality or inequality constraints, we establish global exponential convergence by leveraging contraction theory. Finally, we provide various numerical examples to illustrate the utility of SPPID.

Most studies of stochastic search processes with resetting focus on spontaneous resetting events that occur at a random sequence of times typically generated by a Poisson process. One of the major consequences of spontaneous resetting is that it yields a finite mean first passage time (MFPT) in unbounded domains, resulting in an optimal resetting rate that minimises the MFPT. An optimal resetting rate can also occur in bounded domains provided that the reset point is not too far from the target. Recently, there has been growing interest in event-based rather than spontaneous resetting, where resetting is triggered when a specific threshold is reached. Identifying the threshold with a physical non-target boundary then leads to so-called boundary-induced resetting. In this paper, we explore the effects of boundary-induced resetting on single-particle diffusive search in a two-dimensional (2D) bounded domain $\Omega$ containing one or more small interior targets. The interior target boundaries are totally absorbing, whereas the exterior boundary $\partial \Omega$ is sticky. That is, whenever the particle reaches a point on $\partial \Omega$ it remains attached for a random waiting time $\tau$ after which it immediately resets to a fixed interior point $\x_0\in \Omega$. Using renewal theory, matched asymptotic analysis and Green's functions we calculate the unconditional MFPT for absorption by any of the targets and compare this with the corresponding MFPT of a search process on the same domain with a reflecting surface $\partial \Omega$ and no resetting. We show that boundary-induced resetting only reduces the MFPT relative to the no resetting case if the searcher's reset point $\x_0$ is sufficiently close to one of the targets.

This study derives the asymptotic conditions under which the massless stretched string formulation is valid for the equivalent dynamics of the inertial string carrying a moving point mass. Consequently, the governing equation under such conditions can be written as a singularly perturbed partial differential equation when the inertia of the string is included. When the inertia of the string is ignored, the governing equation reduces to a non-homogeneous hypergeometric ordinary differential equation governing the transverse motion of the moving mass. This hypergeometric equation has a discontinuous solution at the terminating boundary known as the trajectory paradox. However, numerically solving the actual equation yields a continuous solution with a rapid variation near the boundary. The complete solution resembles the inner solution as well, whereas the hypergeometric ordinary differential equation leads to an outer solution, revealing that the trajectory paradox is a boundary layer.

Context. Nuclear star clusters (NSCs) and supermassive black holes (SMBHs) often coexist at the centers of galaxies, engaging in complex dynamical interactions that influence SMBH growth. Stars that approach the SMBH too closely are tidally disrupted, and a fraction of their debris is accreted onto the black hole. Such events may contribute to SMBH growth in gas-deficient galaxies. Aims. Our work aims to improve the theoretical framework for understanding the mass accretion of stellar debris following tidal disruption events (TDEs) through numerical simulations. We perform direct N-body simulations of an NSC with a central SMBH embedded in a fully self-consistent dynamical model to test a new TDE accretion prescription in a set of pilot simulations. Methods. The previous TDE prescription used a fixed accretion radius and assumed that the entire stellar mass was accreted after disruption. In the new partial-mass accretion model, the mass added to the SMBH depends on the orbital eccentricity and penetration factor of the disrupted star. Results. We find that most TDEs involving main-sequence stars are marginally eccentric or marginally hyperbolic and contribute approximately half of the stellar mass to SMBH growth, consistent with recent hydrodynamical simulations. In all simulated models, the TDE-driven mass accretion rate exhibits an early peak followed by a rapid decline. Models with higher initial SMBH masses exhibit a faster decline in the accretion rate, consistent with more rapid loss-cone depletion and inefficient replenishment

Knowing where species occur is fundamental for biodiversity research and conservation. Species distribution models (SDMs) link species observations to environmental conditions to estimate their spatial distribution. However, accuracy varies with the underlying data and models, making it essential to know for which species models can be trusted. Deep-learning-based SDMs ("DeepSDMs") now jointly model thousands of species, drawing on hundreds of millions of community-science records. At this scale, averaging performance hides substantial species-level variability, particularly for rare species, often of greatest conservation concern. Records are also strongly biased, making occurrence counts misleading. Accounting for these factors is essential for a reliable and informative evaluation of multi-species SDMs. Here, we introduce a Sampling-Aware Global Evaluation (SAGE) benchmark, combining GBIF records for training with sPlotOpen vegetation plots for presence-absence evaluation across 5771 plant species. We propose an evaluation framework that groups species based on two properties, sampling effort and relative prevalence, which describe how densely a species' range is sampled and how frequently the species is recorded. Evaluating single-species SDMs and multi-species DeepSDMs, we find that Random Forests and DeepSDMs perform best overall, but neither dominates: DeepSDMs outperform single-species SDMs for infrequently recorded species while offering no consistent advantage for well-sampled ones. Crucially, this advantage emerges only when established bias-correction practices, such as spatial thinning and reweighting, are carried over to the deep-learning setting. SAGE helps identify the species and data conditions for which a given approach is beneficial, thereby supporting the development of more transparent and ecologically credible SDMs. Data and code: https://earens.github.io/sage/

We develop a permutation-invariant Set Transformer to reconstruct the equation of state (EoS) of dense matter from variable-size, unordered sets of neutron star (NS) observations. The model takes stellar masses together with radii, tidal deformabilities, or both, and predicts either the pressure $P(n)$ or the sound speed $c_s^2(n)$ on a fixed density grid, along with density-dependent uncertainties. Nothing in the architecture prescribes which star informs which density: self-attention couples all observations nonlinearly, and each density point reads the full set through its own learnable query, so the star-to-density mapping is learned from the data. Trained on independent piecewise-polytropic and Gaussian-process EoS ensembles, the model provides well-calibrated predictions whose uncertainty increases in density regions that stable stars cannot probe. Reconstruction errors decrease with the number of observations, while tidal deformability generally improves accuracy at a fixed observation count, even when it carries its own measurement noise. Sensitivity analysis reveals a density-local mapping: in the pressure models, predictions at density $n$ depend most strongly on stars whose central densities are near $n$. We also show that the sensitivity of the model to the inferred stellar compactness provides information on the minimum central density. These results demonstrate that set-based neural inference, in which the star-to-density mapping is learned rather than assumed, can extract physically interpretable EoS information with calibrated uncertainties.

We investigate the processing of the circumstellar polycyclic aromatic hydrocarbons (PAHs) associated with young stellar objects (YSOs) through their stellar activities. We analyze the AKARI near-infrared (2.6$-$4.4 $\mu$m) long-slit spectral data of 10 YSOs, focusing on the aromatic feature at a wavelength of 3.3 $\mu$m and the aliphatic features at 3.4$-$3.6 $\mu$m. The spectral fitting results show large variations from target to target in the flux ratio of the aliphatic to aromatic features, the shape of the aliphatic feature profile, and the central wavelength of the aromatic feature. In particular, HD 97048 and Elias 1 show strong aliphatic features at wavelengths longer than 3.5 $\mu$m, which is consistent with the hydrogenated nanodiamond features reported in previous studies. We find that the properties of the hydrocarbon dust vary under the influence of the YSO activity, where the degree of dust processing is likely to be related to the age of the YSO and the amount of the circumstellar dust. In addition, correlation is observed between the variations of the H$_2$O and CO$_2$ ice features and the aliphatic feature profiles, suggesting connection between the formation of CO$_2$ ice and the processing of aliphatic hydrocarbons. Finally, we find a significant correlation between the variations in the central wavelength of the aromatic feature and the shape of the aliphatic feature profile. These findings may be consistent with a tentative scenario in which circumstellar PAHs are converted to nanodiamonds through irradiation by high-energy particles produced by YSO activity.

Rare-earth orthoferrites such as DyFeO$_3$ exhibit a rich interplay between localized rare-earth and transition-metal moments, giving rise to complex magnetic phases and magnetoelectric phenomena. Understanding their electronic and spectroscopic properties from first principles requires an accurate description of the localized Fe-$3d$ and Dy-$4f$ states and their hybridization with O-$2p$ states. Here, we combine first-principles calculations and x-ray absorption near-edge structure (XANES) measurements to investigate the electronic structure and the O $K$-edge spectrum of DyFeO$_3$. We employ density-functional theory (DFT) with Hubbard $U$ corrections (DFT+$U$) determined from first principles using density-functional perturbation theory, the HSE06 hybrid functional, and orbital-resolved DFT+$U$ with Hubbard parameters calibrated to reproduce the electronic structure computed using HSE06. We find that standard DFT+$U$, despite improving the band gap, substantially underestimates the crystal-field splitting of the unoccupied Fe-$3d$ states and consequently fails to accurately reproduce the separation of the two lowest-energy features in the O $K$-edge spectrum. HSE06 provides a more balanced description of the relevant electronic states, including the Fe-$3d$ crystal-field splitting. Mapping the corresponding electronic structure onto orbital-resolved DFT+$U$ yields a spectrum that remarkably reproduces the two lowest-energy experimental features and captures the main characteristics of the spectrum at higher energies. These results establish the low-energy O $K$-edge features as a sensitive probe of the Fe-$3d$ crystal-field splitting, while showing that the localized Dy-$4f$ states leave no distinct spectral fingerprints despite their importance for the magnetic properties.

Debates amongst agents driven by large language models (LLMs) have demonstrated vast potential in various applications, but when these interactions include humans and take place in high-stakes environments, e.g., in public policy deliberations, they are beset with issues such as sycophancy and a lack of faithful explanations. To tackle these issues, we present OmouAI, an interactive and inclusive deliberation system that uses LLMs in combination with computational argumentation, a field which excels in representing and reasoning within debates. OmouAI allows a human user to deliberate policy claims for real-world challenges with simulated personas, e.g., representing stakeholders, domain experts or devil's advocates, towards reducing sycophancy. Each persona generates its own arguments, and the arguments of all parties form a shared argumentation framework. Users can then contest, add and revise arguments, providing crucial human oversight. Then, arguments are evaluated using deterministic argumentative semantics against external goals, such as the UN Sustainable Development Goals, guaranteeing faithful explanations. The advancement or worsening of the goals thus serve as indicators for the policy recommendations.

Sidechain packing is a critical stage in protein folding, with direct implications for structure-based drug discovery. Google DeepMind's tool, AlphaFold2, predicts protein backbone reliably but sidechain positioning less accurately. Recovering the lowest-energy rotamer assignment over a fixed backbone is NP-hard. We present a hybrid quantum-classical pipeline that repacks sidechains on the AlphaFold backbone using the Quantum Approximate Optimisation Algorithm (QAOA), encoding the one- and two-body energies as a quadratic unconstrained binary optimisation (QUBO) problem. We introduce a constraint-preserving ansatz, pairing a W-state initialisation with a cyclic XY ring mixer, that enforces one-hot rotamer validity without penalty terms while keeping two-qubit gate scaling linear in the rotamer count. We also define an asymptotic shot-scaling metric, measured against the experimentally resolved conformation, that fixes optimiser quality independently of the baseline; its fitted growth stays below the classical exhaustive-search rate at moderate rotamer flexibility. Evaluated on bovine pancreatic trypsin inhibitor (5PTI) across high- and moderate AlphaFold-confidence regions, the pipeline lowers conformational energy against the AlphaFold baseline.

Language agents struggle to act and learn in environments that require long sequences of low-level actions. Code-based abstractions can make these agents more productive by letting them invoke reusable skills instead of repeatedly selecting individual actions. The code handles recurring local decisions, while the language model decides which skills to use and how to combine them. Yet abstractions are leaky, and situations beyond a skill's capabilities may require a return to primitive actions. Motivated by this tradeoff between productivity and flexibility, we systematically study how code-based action abstraction affects the performance, inference cost, and learning of language agents. We study this in NetHack, a challenging, long-horizon game environment, using CodeHack, our library of code-based skills with natural-language descriptions. We use this library to compare agents restricted to primitives with those using semantic skills alone or in combination with primitives. We evaluate these agents in three settings: zero-shot prompting, supervised fine-tuning, and reinforcement learning. Across a broad zero-shot evaluation on NetHack, we find that compared with primitives, skills nearly triple game progression, while reducing inference cost per episode by 86%. Combining skills with primitives retains much of this benefit while preserving a path back down to low-level actions. Finally, in RL, we find that skill-based agents learn significantly faster than agents acting on primitives, achieving a 7.2x larger average gain in dungeon level over the same training budget. These results show that a supplied skill library can improve performance, efficiency, and learning, while retaining primitives provides flexibility when the library is insufficient. We release CodeHack together with training and evaluation code.

Among the various "no-go theorems" in the foundations of quantum mechanics, Bell inequalities are among the few, which do not implicitly rely on the mathematical structure of quantum mechanics. Instead, Bell inequalities are commonly understood to rest on Kolmogorov's theory of probability and a "locality" assumption. Physical evidence supporting their violation is hence taken as evidence against at least one of these two assumptions. This work provides a comprehensive analysis of the physical debate on Bell inequalities from the point of view of Kolmogorov's theory of probability, focusing on the Bell-Clauser-Horn-Shimony-Holt inequality (BCHSH inequality). The central contribution is the motivation and mathematical study of the notion of contextuality in Bell experiments: in so-called contextual Bell models the detectors themselves behave randomly and the distribution of the random detector parameters depends on the chosen detector settings. It is shown how even "classical" Bell experiments motivate such a description. The need to account for contextuality was pointed out before by de la Peña, Cetto, Brody, Lochak, Bohm, Hiley and others. A broad class of contextual models is proven to satisfy the BCHSH inequality, thereby generalizing a previous result by Gill and Lambare. Nonetheless, even under strong locality requirements, contextual Bell models may violate the BCHSH inequality: based on work by Popescu and Rohrlich, we consider a "proof of concept" example exhibiting the maximum possible violation. Contrary to the much debated "loopholes" in empirical tests of Bell inequalities, contextuality thus reveals a conceptual misunderstanding in the physical conclusions commonly drawn from Bell inequality violations.

Resource constraints make high-dimensional hyperspectral imaging challenging in autonomous perception, motivating the use of band selection methods. However, the sensitivity of band-selection methods to sampled data and their relationship to semantic segmentation models (SSMs) remain underexplored. This study evaluates six band selection methods on ten independently sampled, class-balanced region-of-interest (ROI) sets, yielding 60 top-25 band subsets from the Hyperspectral City V2 (128 bands: 450-950nm) dataset. Top-$K$ bands ($K\in\{3,5, ... 13\}$) from the first three ROI sets are evaluated with three SSMs against the corresponding 128-band baseline. Experiments show that intra-method stability is method-dependent: Sim-LP shows the highest stability (pairwise Jaccard similarity) and, together with JMIM+CSNR, yields the best segmentation results. Top-$K$ based SSMs remain competitive with baselines, with gains of up to 2.01 mIoU and 1.72 mF1 points, and 18-22x faster CPU inference for $K=9$. However, performance does not improve monotonically with $K$, and stability shows no consistent association with SSM performance. These findings suggest that intra-method stability is informative but an unreliable indicator of downstream segmentation performance, highlighting the need to evaluate band-selection methods across repeated samples, subset sizes, and SSMs.

Muon flux variations in muography time series can provide information on density-related changes in geological targets. A common approach is to define regions of interest before constructing flux time series, but this choice may dilute coherent signals or mask localized variations. We present a data-driven analysis of two muon data acquisitions at the Sos Enattos Mine in Sardinia, Italy. The method aggregates lines of sight into overlapping spatial blocks, computes standardized flux deviations, and applies principal component analysis through singular value decomposition. The leading components reveal spatially coherent structures and distinct temporal evolutions. In each run, the first mode explains more than 20% of the variance, while the first eight modes explain approximately 80%. Regions selected from the spatial modes show anticorrelated and delayed flux variations. These results show that the proposed approach can identify candidate regions of coherent muon-flux variability without imposing them a priori.

We present a direct simulation Monte Carlo (DSMC) method for the kinetic theory of ordered fluids recently introduced by the authors. The microscopic constituents of these fluids carry an order parameter belonging to a manifold. The method applies Strang splitting to couple a Monte Carlo treatment of the binary collisions with a symplectic integration of the mean-field Vlasov transport on the order-parameter manifold. We use the Nanbu-Babovsky approach for Maxwellian collision kernels, and Bird's acceptance-rejection method for general ones. The mean-field Vlasov force is reconstructed from the particle ensemble at each time step. We also include a Bhatnagar-Gross-Krook (BGK) / Andersen thermostat to control the system temperature. We formulate the method for a generic order-parameter manifold, and specialize it to two-dimensional calamitic (rod-like) molecules. Numerical tests validate the method. We confirm the relaxation to the Maxwellian distribution predicted by the H-theorem, and simulate the mean-field alignment driven by quadratic, Kuramoto, and Onsager interaction potentials. With the Onsager potential we successfully capture the isotropic-nematic phase transition at the correct critical temperature.

Corr2Cause asks whether a causal claim holds in every DAG compatible with observed correlations and conditional independencies. We frame this as latent-object reasoning: the label is defined by a CPDAG query, but free-form chain-of-thought often collapses the Markov-equivalence-class problem into local pattern matching. We propose Structured Thinking, a two-turn pipeline that first externalizes a typed, schema-constrained CPDAG summary and then answers against that graph state. On the Corr2Cause full test, Structured Thinking raises Qwen3.5-27B from $73.0$ to $86.4$ $F_1$(Yes) over a strong PC-instruction baseline in the primary paired run ($+13.4$ pp; McNemar $p=2.4\times 10^{-6}$; bootstrap $95\%$ CI [$+8.4$, $+18.6$]); across three full-ID seeds, the mean gain is $+8.1 \pm 5.3$ pp. A PC-scaffolded two-turn prose control reaches only $67.6$ $F_1$, indicating that a detailed PC scaffold plus a schema-free prose intermediate is not sufficient. The same pattern holds on Qwen3.6-27B, Paraphrase-OOD, and GPT-5.4-mini. Scrambling the emitted CPDAG costs $12.0$ pp $F_1$, and a full-split audit shows close agreement with the reference CPDAG (ID skeleton $F_1$ $0.960$; exact match $75.9\%$). These results support a bounded design principle: externalize the latent object that defines the label, constrain its form, and test whether downstream answers use it.