Papers

Evaluating AI agents on scientific research tasks is constrained by the time and resources required for the underlying experiments or calculations. In computational materials research, repeating the same expensive simulations across agents and trials can make evaluation impractical. We introduce CompMat-Bench, a benchmark of 94 tasks derived from recently published computational materials studies, each asking agents to complete a step toward achieving the study's scientific goal. We reproduce the research steps in advance and assess agents on preparing inputs and analyzing outputs for expensive simulations, so expensive simulations can be avoided during evaluation. The reproduced inputs and results serve as ground truth for grading agents with fixed rules, without an LLM judge. The benchmark supports four evaluation conditions: single tasks and workflows composed of related tasks, each with full or reduced methodological guidance. With full guidance on single tasks, agents based on three LLMs demonstrate the ability to complete individual materials research steps, with pass rates of 66.0-90.4% across 94 tasks. Both longer workflows and reduced guidance can limit agent performance, but in different ways for different agents: they lower the pass rates of the weaker agents, whereas the strongest agent falls only when a long workflow is combined with reduced guidance. Failure analysis attributes most failures to scientific errors rather than to errors in software usage. CompMat-Bench provides a basis for comparing agents on the steps of real materials research and for analyzing agent failure modes.

Non-ergodic quantum many-body dynamics offers a route to persistent quantum coherence far from equilibrium, beyond the conventional expectations of thermalizing statistical mechanics. Most known examples have been identified through intuition, analogy, or numerical search, rather than by systematic Hamiltonian design. This leaves few general methods for engineering robust non-ergodic dynamics in disorder-free interacting quantum many-body systems, especially in higher dimensions where enhanced connectivity generally disfavors ergodicity breaking. Here, we use an eigenstate-to-Hamiltonian construction approach to systematically engineer two-dimensional quantum spin Hamiltonians with tunable non-ergodic properties. Starting from a common parent model and input eigenstate, we construct two target Hamiltonians on the square lattice that differ in their interaction geometry: axial versus diagonal. We show that the resulting interaction geometry and coupling pattern control the Hilbert-space structure and dynamics. The axial model forms a connected Hilbert-space network and exhibits an anisotropy-driven crossover from ergodic to non-ergodic behavior through the suppression of resonances. In contrast, the diagonal model exhibits Hilbert-space fragmentation and, within its largest irreducible sector, signatures of quantum many-body scar dynamics including towers of highly localized, low-entanglement eigenstates, and long-lived coherent revivals. These results establish a direct link between Hamiltonian design, Hilbert-space network structure, and emergent non-ergodic dynamics.

Activity can enhance the long-wavelength density fluctuations of a solid beyond the equilibrium fluctuations predicted by the Mermin-Wagner-Hohenberg theorem, but the microscopic origin of this enhancement remains unclear. We study a two-dimensional active Brownian crystal whose longitudinal and transverse displacement covariances scale as $C^u_\lambda(q)\sim q^{-3}$ as compared to the passive counterpart with $q^{-2}$ law. This spectrum predicts an MSD-plateau divergence proportional to $L$ in two dimensions, logarithmic in three dimensions, and finite in four dimensions, in close correspondence with recent results reported in Dey et al. [Nat. Commun. 16, 5498 (2025)]. Inverse of the Covariance gives $H^{\mathrm{cov}}_\lambda\sim |q|^3$ with the corresponding dispersion $\omega_{\mathrm{cov},\lambda}\sim q^{3/2}$, whereas direct mechanical response gives $K_{\mathrm{resp},\lambda}\sim q^2$ and acoustic modes with $\omega_{0,\lambda}\sim q$. The nonlinear dispersion is therefore covariance-defined rather than mechanical. The mobility and integrated active-force correlations remain non-singular, although a finite-persistence crossover cannot be ruled out completely with the existing data. Microscopically, the longitudinal and transverse Irving-Kirkwood interaction-stress spectra scale as $q^{-1}$. The transverse stress-stress correlation connects well with the anomalous displacement field, indicating that activity anomalously populates ordinary acoustic modes, while the conservative interaction network transmits the resulting fluctuations as long-ranged stress. A three-dimensional active FCC crystal exhibits the same paired low-$q$ trends.

The goal of this paper is to provide a rigorous methodology for enforcing database constraints using MS VBA event-driven procedures for software applications built on top of MS SQL Server databases, in the framework of the software engineering Database Constraint-Driven Design and Development. The original contribution is a pseudo-code algorithm for assisting developers in this process. We exemplify the results of using it with some VBA code examples taken from a genealogy database software application designed and developed using this methodology. Consequently, it enforces all the database constraints -be they relational or not- which govern this sub-universe of discourse, thus guaranteeing the highest possible quality of its managed data. We also show that the proposed algorithm works fine not only for other relational database management systems, but even with NoSQL platform backends, and that the modifications needed to adapt it to other similar SQL embedding frontend platforms are minimal.

Initially discovered optically, Tidal Disruption Event (TDE) AT2018hyz was first detected in radio 972 days later. The continued monitoring of the event revealed a late-time (1370-2160 days) radio brightening. The source for this radio brightening has been associated with an afterglow originating from an off-axis jet or a delayed mildly relativistic outflow. In this work, we explored the conditions required for a delayed off-axis jet to explain the observed radio brightening. We find that the combined effect of different delay times (t_delay) and observer angles (theta_obs) can produce radio afterglow, as observed for AT2018hyz. The further off-axis from the jet core an observer is, the shorter is the delay inferred between the optical disruption and launching of the jet. This presents a degeneracy between different theta_obs and tdelay. We estimate the sky projection of the jet for these different combinations to break this degeneracy. The spread in sky images ranges from 0.6 mas for a 30 deg off-axis observer to 0.3 mas for an observer perpendicular to the jet core. Future VLBI observations should be able resolve these structures. Thus a combined observation of the light curve and sky images for AT2018hyz (and other jetted TDEs) can be used to narrow down the observer angle and any delay in launching of the underlying jet.

We formulate the corner transfer matrix equations, and the corner transfer matrix renormalization group built on them, as a set of local equations on general planar graphs. These equations depend only on the local subgraphs of the planar graph and are glued together consistently to describe the whole graph. Solving them yields message tensors and corner matrices, which can be used to construct an MPS environment for any connected subgraph. We test the proposed equations on the ferromagnetic Ising model, on a variety of infinite lattices (including hyperbolic ones) and on finite graphs, and find systematic improvement over belief propagation.

We present PLACE, a method that extends the pretrained any-to-audio model AudioX for binaural generation from arbitrary combinations of text, video, and optional audio prompts. PLACE augments video conditioning with Perception Encoder Core features, aligns text and video representations to derive spatial cues, and applies a conditioning-dependent low-rank transformation to the generated latent. The adapter is supervised via decoded-audio interaural level and time difference objectives. Trained on MRSAudio, PLACE improves most metrics over ViSAGe on FAIR-Play and achieves an improved SpatialCLAP score over SpatialSonic on the BEWO-1M Single Static test split. Listener evaluations favor PLACE for video-to-audio and out-of-distribution text-to-audio generation, demonstrating flexible multimodal control and improved spatial consistency.

The integration of Large Language Models (LLMs) into multi-agent systems has shown great potential for automated debugging. Yet nearly all current frameworks rely on rigid, predefined architectures: the number of agents, their roles, and their interaction patterns are fixed before any analysis of the bug occurs. This one-size-fits-all approach is fundamentally mismatched to the heterogeneous nature of software defects. Simple bugs waste resources on unnecessary coordination, while complex ones suffer from insufficient or poorly aligned expertise. This paper introduces ASAD, an adaptive agentic system for debugging that configures its team according to the nature and complexity of each bug. ASAD initiates the debugging process by analyzing the faulty code and dynamically determines the number of agents to deploy, the specialized roles they should have, and the collaboration strategy they should follow. A central coordinator orchestrates this process through iterative planning, reflection, and execution; applying fast single-pass repairs for simple issues while assembling purpose-built teams to tackle more complex failures. We evaluate ASAD on three established benchmarks: Defects4J, DebugBench, and CodeFlaws, using multiple LLMs, such as DeepSeek-V3, Qwen-3 and GPT-5. ASAD consistently improves bug-fix rates by 12–20% over chain-of-thought(CoT) prompting and consistently outperforms static multi-agent systems by 4–9% in fix precision while reducing average agent usage by 32%. Crucially, our system dynamically adjusts the number and roles of agents: it resolves simple bugs with minimal coordination and scales agent involvement only for more complex cases.

We present three locomotion adaptation approaches: Motion Acceleration, Turn Acceleration, and Motion Deceleration to improve postural stability during body-state transitions in virtual reality (VR). The system detects standing-to-walking, turning, and walking-to-stopping transitions and applies adaptive locomotion smoothing. Motion Acceleration gradually increases locomotion speed when users begin walking, Turn Acceleration smooths rotation while turning, and Motion Deceleration gradually reduces movement speed before stopping. We evaluated these techniques in a virtual navigation task using objective and subjective balance measures. Preliminary results show reduced center of pressure (COP) velocity and improved balance confidence. These findings suggest that locomotion adaptations can improve balance and navigation experience.

We study non-autonomous basins of $F_j$ automorphisms of $\mathbb C^2$ satisfying $$a_j\|p\|\leq\|F_j(p)\|\leq b_j\|p\|\qquad (\forall p\in \mathbb B)$$ where $0<a_j<b_j<b<1$. We show that if $b_j^m<a_j$ for some integer $m$ and all $j>0$ then every bounded plurisubharmonic function on the non-autonomous basin is constant. Moreover, if in addition we have $\prod_{i=1}^{j-1}b_i\leq b_j^m$ then the basin is biholomorphic to $\mathbb C^2$. For the quadratic Hénon family $$F_n(z,w)=\bigl(z^2+c^{q_n}w,c^{q_n}z\bigr), \qquad 0<c<1,$$ we study how the asymptotic behavior of the sequence $q_n\to\infty$ influences the complex structure of the basin. In particular, we identify broad regimes in which the non-autonomous basin is either biholomorphic to $\mathbb C^2$ or is a Short $\mathbb C^2$, thereby isolating possible sequences for which other complex structures may occur.

The microscopic mechanism governing the zero-field incommensurate magnetic order in the honeycomb cobaltate $\mathrm{BaCo}_2(\mathrm{AsO}_4)_2$ remains a significant unresolved problem, particularly concerning the relative contributions of bond-directional Kitaev-type interactions and exchange frustration. This study investigates an extended $J_1\text{--}K\text{--}\Gamma\text{--}\Gamma'\text{--}J_2\text{--}J_3$ model, which incorporates XXZ-type exchange anisotropies on the honeycomb lattice. Nearest-neighbor parameters are constrained by ab initio electronic structure calculations. By integrating the analytical Luttinger-Tisza approach with exact diagonalization calculations under twisted boundary conditions, we delineate the classical and quantum phase diagrams across the $(J_2, J_3)$ parameter space. We demonstrate that the stabilization of the experimentally observed incommensurate spiral phase, propagating along the $\Gamma\to M$ direction, does not necessitate an anomalously dominant Kitaev coupling. Instead, this phase arises naturally from the synergistic interplay between exchange frustration, driven by further-neighbor couplings ($J_2, J_3$), and intermediate off-diagonal bond anisotropies ($\Gamma, \Gamma'$). In the quantum regime, we elucidate the competition between this incommensurate manifold and an out-of-plane ferromagnetic ($\mathrm{FM}_z$) phase, which is selectively stabilized via a quantum order-by-disorder mechanism. Our findings reconcile conflicting interpretations of the magnetic interactions in $\mathrm{BaCo}_2(\mathrm{AsO}_4)_2$ and establish the microscopic origin and stability range of its incommensurate ground state.

We show that the Wasserstein space of the circle is not flat, in contrast with the flatness of the Wasserstein spaces of the real line and of an interval. Its sectional curvature is governed by a symplectic form on the Fourier coefficients, so that a plane is flat exactly when it is isotropic for that form, and on the plane spanned by the $n$-th Fourier modes the curvature is bounded above by $3n^2$, with equality exactly at the uniform measure. It is therefore unbounded. We compute it inside the formal Riemannian structure of Otto and confirm it directly from the definition of $W_2$, which also shows that no neighbourhood of the uniform measure in $P_2(S^1)$ is CAT(0). The result corrects the widely stated expectation that a Wasserstein space is flat whenever its base manifold is one-dimensional. The computation is carried out in the global frame of the tangent bundle of $P^\infty(M)$ induced by the eigenfunctions of the Laplacian of $M$, whose existence was observed by Lott. We prove that it is a Schauder basis of every tangent space and write Otto's metric and its Levi-Civita connection in it. On the circle this gives the metric and connection coefficients of $P^\infty(S^1)$ in closed form, as finite combinations of the Fourier coefficients of the density. A final section puts the frame to work on a computation from dynamics, reading the derivative of the push-forward action of an expanding circle map at its invariant measure as a weighted shift on the Fourier frequencies.

We investigate the distributions of off-diagonal scattering matrix elements and cross-sections in the Ericson regime for time-reversal invariant systems with odd spin. Within the universal Heidelberg approach we prove that Gaussian and exponential behavior is found as the leading-order term of an asymptotic expansion of the aforementioned distributions. Furthermore, the subleading-order term is derived and compared with numerical simulations.

Speculative decoding has achieved substantial lossless speedups for LLMs, but remains less effective for large vision-language models (LVLMs), where lightweight drafters struggle to use rich multimodal information. A second limitation is that standard distillation supervises the drafter only along the original training trajectory, without modeling how target predictions shift after the drafter's own proposals. As drafting moves away from this trajectory, the drafter can increasingly disagree with the target, reducing acceptance in later steps. We propose HAWK to address both limitations. HAWK uses representation similarity to select informative target layers and learns how to combine their hidden states. For visual information, it directly provides the drafter with compressed visual hidden states from the target model instead of raw visual tokens, making the visual information easier for a shallow drafter to use. HAWK also trains the drafter to capture how target predictions change after its own proposals, improving its agreement with the target during multi-step drafting. On SmolVLM-256M across ten multimodal benchmarks, HAWK raises average acceptance length from 3.32 to 4.08 and speedup from 2.19x to 2.60x over EAGLE-3 under greedy decoding, and from 2.89 to 3.41 and 1.92x to 2.19x under sampling.

Software practitioners increasingly rely on Large Language Models (LLMs) to generate code that integrates external libraries. However, LLMs often produce incorrect library usage, such as invalid imports, outdated API calls, and hallucinated dependencies, leading to compilation or runtime failures that reduce the reliability of AI-assisted software development. In this paper, we propose an agentic approach to mitigate libraryrelated errors in LLM-generated code. More specifically, we first conduct an exploratory study to characterize the library-related issues produced by LLMs. Our analysis of 100 LLM-generated code files reveals that 84% of generated files contain at least one library-related error, with recurring patterns including incorrect import paths, missing imports, hallucinated libraries, deprecated library usage, and unused imports. Based on these findings, we design an agentic approach that integrates task analysis, documentation grounding, code generation, and automated validation to improve library usage during code synthesis. We evaluate our approach on 300 code generation tasks derived from realworld implementations of rapidly evolving Python frameworks, including LangChain and AutoGen, across five LLMs: GPT-5, DeepSeek-V3, Qwen3, Mistral, and Llama 3. The results show that our approach consistently improves code generation quality across all evaluated models, reducing library-related errors by 38.1% - 54.6% and increasing code correctness by up to 16%.

Mixture modeling is a long established machine learning technique for learning large sets of multi-modal data. While it is known that sequence-to-sequence models for dialog response generation suffer from the problem of low diversity, we hypothesize that it is because sequence-to-sequence models tend to learn a degenerate uni-modal distribution of responses. We then propose to incorporate a mixture of decoders into sequence-to-sequence models and try to make each decoder learn specialized topics in order to improve the diversity of generated responses. Our model is developed under the framework of conditional variational autoencoder (CVAE). We evaluate our approach on an open domain chat corpus and show improvement over strong baselines in quantitative measures and human evaluation.

Grokking refers to the delayed emergence of validation-set generalization after a model has already overfit the training set. Although first observed in small algorithmic tasks trained with transformers, its underlying mechanism remains unsettled. In this work, we develop an analysis framework based on mode connectivity and the geometry of low-loss regions. The framework predicts that the standard modular-arithmetic setting does not always produce grokking: under a symmetry-preserving train/validation split, we observe a stable anti-grokking case in which validation performance does not recover. This counterexample challenges several existing correlational explanations of grokking. More broadly, our analysis framework and results further suggest that grokking arises when the low-loss regions induced by the training and validation partitions are misaligned. Once these regions become well aligned, training hyperparameters alone cannot produce grokking and the observed dynamics collapse to either trainable or non-trainable behavior.

In this paper, we extend earlier findings of supra-competitive outcomes in optimal-execution games by identifying a learned punitive mechanism that deters deviations and provides behavioural evidence of collusion. We investigate this mechanism in a two-player, finite-horizon Almgren-Chriss liquidation game. Independent proximal policy optimisation agents with access to within-episode price and action histories achieve costs below the Nash benchmark. We identify a profitable deviation by training against the mean learned liquidation schedule, then impose its first trade on one of the original agents. The opponent responds by accelerating liquidation. This response more than offsets the deviator's gain in every run and both player roles, while leaving the punisher's average payoff materially unchanged relative to not punishing under the same deviation. The punisher imposes greater losses on the deviator while preserving its own average payoff, despite the availability of more profitable, less punitive liquidation plans. Matching deviations and subsequent additional selling rise and later decline during training, while final policies retain an effective punitive response. We formalise two checks: whether punishment outweighs the gain from deviating, and whether the change in trading behaviour is large enough to account for the loss imposed. Both checks hold for the tested deviation. Together, these findings provide behavioural and economic evidence supporting a collusive interpretation of the learned supra-competitive outcomes.

Wave-based coherent imaging, including terahertz tomography, synthetic-aperture acoustics, and millimeter-wave radar, forms images by Fourier-processing finite-length signals, with an exact point spread function that is not Gaussian but a Dirichlet kernel: complex-valued, oscillatory, and periodic. However, transplanting 3D Gaussian splatting to coherent sensing fails by construction; Gaussian splats discard the sidelobe energy (10-20% of the total) and the phase that governs coherent interference between reflectors. Our key idea is to replace the learned Gaussian footprint with the physically exact Dirichlet kernel of the finite-window DFT, modulated by a surfel that carries area, normal, and material, so that the rendering primitive matches the measurement physics instead of approximating it. We pair this primitive with a specialized solver, Dirichlet Sliding Frank-Wolfe (DSFW), that combines variable projection, residual dual certificates, and certificate-driven hard replacement of low-utility surfels, with periodic low-resolution coupled Levenberg-Marquardt correction, navigating the rugged loss landscape that breaks generic first-order optimizers. The Dirichlet kernel admits an O(1) closed-form evaluation, so the forward model matches FFT ground truth to machine precision while remaining differentiable end-to-end. On dense terahertz reconstruction, our method recovers reflector centers to 0.018 bin RMSE, 10-50x faster than waveform-level automatic differentiation, where Gaussian splats fail.

Dynamic borrowing can improve efficiency in clinical trials using external information, but existing methods may not offer transparent, prespecified control over borrowing. We introduce Standardized Dynamic Borrowing (SDB), which combines a source-informed component with a no-borrowing component and updates their relative odds using a Bayes factor standardized by its supremum over admissible target data. The posterior density under the no-borrowing component is obtained by normalizing the target-data likelihood. The prespecified source-component initial weight is a sharp uniform upper bound on its posterior weight, providing an interpretable borrowing constraint. When source and target models belong to the same one-parameter canonical exponential family, posterior expected local-information-ratio effective sample size equals target sample size plus borrowed source-prior effective sample size minus a nonnegative heterogeneity penalty. For normally distributed outcomes, borrowing depends on a standardized source-target conflict $Z$-score and decreases according to a standard normal kernel. We also jointly select the initial weight and target sample size to meet prespecified power and Type I error requirements. Normal and binomial illustrations demonstrate how SDB translates source-target compatibility into data-adaptive borrowing and how the calibration procedure prespecifies the source-component initial weight and target sample size under operating-characteristic constraints. These results provide a transparent framework for prespecifying, controlling, and interpreting dynamic borrowing.

We examine how local motif structure and global network topology jointly shape spiking dynamics in stochastic neuronal networks. Using networks of Izhikevich neurons with Erdős-Rényi (ER) and scale-free (SF) background connectivity, we compare motif-embedded networks with synapse-count-matched, non-motif controls under noise- and stimulus-driven protocols. Motif embedding increases noise-induced coherence in both topologies and provides a smaller improvement in signal transmission, while SF-based networks attain greater absolute coherence and transmit signals faster and more reliably than ER-based networks. Motif-rewiring experiments show that this benefit depends on the specific arrangement of intra-motif connections rather than on strong local connectivity alone: randomizing the motif skeleton reduces coherence overall, although rewiring the non-recurrent feed-forward loop and bi-parallel motif can increase it. Targeted ablation shows that the SF advantage depends in part on hub integrity; removing high-degree excitatory neurons impairs coherence and signal transmission more than matched random node or edge removal. Together, these results indicate that local motifs and global hubs make distinct, complementary contributions to coherent spiking dynamics.

One might think that learning the identity function with a deep linear residual network is trivial - the path along residual connections already implements the identity, and so the network need only drive its weights to zero. However, this zero-weight solution is just one point on an entire manifold of population-loss minimizers, each corresponding to a different decomposition of the identity across the network's layers. Although the population loss does not distinguish among these solutions, stochastic gradient descent (SGD) reproducibly favors particular ones. For instance, under anisotropic label noise, the learned layers exhibit a noise-dependent spectrum; even with weight decay, SGD does not generally recover the zero-weight solution. Changing only the parametrization, while leaving the set of realizable functions unchanged, yields different behavior: factoring each weight matrix as a product of two matrices causes the weights to collapse to zero, even without explicit weight decay. While perhaps mysterious and unintuitive at first, these phenomena can be understood through the lens of entropic loss, which augments the population loss with a term proportional to the expected squared norm of the minibatch gradient (Ziyin et al., 2025). On the identity manifold, the population loss is constant, while the entropic term distinguishes among these decompositions. We characterize its minimizers analytically and use them to derive predictions for the structure of solutions favored by SGD. Networks trained with SGD closely match these predictions. Overall, the identity learning task studied here serves as a clean and simple case study of how the lens of entropic loss can clarify why SGD favors particular decompositions of the same input-output function.

We study total path length in recursive trees with positive deterministic attachment weights. Writing $W_n=\sum_{i=1}^n w_i$ and $p_n=w_n/W_n$, we obtain exact martingale-innovation identities and a variance recurrence. Under the condition $p_n=O(n^{-1})$, centered total path length divided by $n$ converges almost surely and in $L^2$ to a nondegenerate random variable, and its variance is asymptotic to a positive constant times $n^2$. No polynomial asymptotic for $W_n$ is required. For power weights $w_i=i^\alpha$, the same argument applies to every real $\alpha$, including the critical and summable regimes beyond the positive-power cumulative-weight assumptions of existing profile theory. The expected average depth is logarithmic for $\alpha>-1$, iterated logarithmic for $\alpha=-1$, and bounded for $\alpha<-1$, while the global fluctuation scale remains linear throughout. In the summable regime we identify the random limit through the weighted depths of the infinite tree. The uniform case recovers the classical variance coefficient $2-\pi^2/6$. For linear weights we evaluate the coefficient as $8-2\pi^2/3$. Although this tree and a random binary search tree have identical insertion-depth marginals and expected total path length, their asymptotic variance coefficients differ by one. This gives an explicit comparison of global dependence that is invisible in individual depth distributions.

Large language model (LLM) based agents are often criticized for lacking spatial understanding and mainly exploiting statistical text patterns. We investigate their spatial comprehension through an architecture combining geometrical tools with a LLM serving as a high-level orchestrator in grid-world environments. The agent first collects geodesic trajectories, which are then vector-quantized to extract a representative subset. Offline, the LLM associates a natural language description of the underlying behavioral patterns to each selected trajectory, making it a tool. Online, the LLM chooses the appropriate tool conditioned on the current state and goal. Low-level control is handled by primitive actions that execute the trajectory associated with the tool. From an agentic AI perspective, this approach separates learning into two levels: tool discovery is handled through unsupervised quantization of trajectories, while reasoning and decision-making are handled by the LLM. We test the approach in a partially observable dynamic 2D grid environment with an open vision-language model (Qwen3.6-35B-A3B). Pairing the geometry-derived tool library with an agent-centered zoom tool and a collision detection tool lets a fast, non-reasoning configuration match the goal-reaching rate of a much more costly chain-of-thought version, while cutting the cost of a decision from minutes to seconds.

We present the first detailed X-ray and optical study of the intermediate polar Swift J0614.0+1709 based on a pointed XMM-Newton observation complemented by long-term optical photometry from TESS. A strong coherent modulation at the white dwarf spin period was detected in X-rays, with a period of $\sim$1411 s, confirming the magnetic nature of the system. The X-ray spin pulse exhibits a pronounced energy dependence, with the modulation amplitude decreasing towards higher energies, consistent with phase-dependent absorption in the accretion flow. Phase-resolved spectroscopy further shows that the spin variability is primarily driven by changes in the covering fraction of the partial-covering absorber. The broadband X-ray spectrum is well described by a partially absorbed thermal plasma model with a temperature of $\sim$12 keV and a soft blackbody component at $\sim$63 eV. The inferred bolometric luminosity of $\sim$5$\times$10$^{34}$ erg s$^{-1}$ places the source among the more luminous intermediate polars. The TESS observations reveal substantial sector-to-sector variability in the relative strengths of the orbital and spin modulations. While the spin signal remains persistently dominant, the beat modulation varies significantly and disappears entirely in one Sector, suggesting temporal changes in the contribution of the reprocessing region(s). The orbital and, to a lesser extent, spin modulation amplitudes show some tendency to be stronger during brighter optical states, suggesting that changes in the mass transfer rate may influence the optical emission and modulation amplitudes. Overall, Swift J0614.0+1709 is a comparatively X-ray luminous intermediate polar exhibiting complex optical variability characterised by changes in pulse morphology and recurrent brightening events, making it a valuable target for future simultaneous optical and X-ray studies.

This work details some of the recent radiological studies performed at National Synchrotron Light Source II (NSLS-II) using the Monte Carlo (MC) particle transport code FLUKA. In particular, it assesses the radiation protection challenges posed by the anticipated NSLS-II facility upgrade, taking into consideration various upgrade options. Also, it evaluates the impact related to the use of different shielding materials and the employment of ambient dose H* as the external radiation exposure estimator of effective dose. Lastly, a demonstration of synchrotron radiation transport is provided for an in-silico beamline, all the way from the radiation source to end-station, utilizing FLUKA's recent X-ray reflectivity capabilities.

Explainable suicide-risk assessment requires models not only to estimate risk severity, but also to identify supporting language and the risk and protective factors expressed in a post. We present our system for the IEEE BigData 2026 Cup on Explainable Suicide Risk Assessment on Social Media, which addresses three tasks: risk-level classification, evidence phrase extraction, and multi-label factor identification. Our approach adapts Qwen2.5-Instruct models using quantized low-rank adaptation (QLoRA) and an answer-masked causal language-model objective. We jointly train across all three tasks for risk classification, jointly train on Tasks 1a and 1b for evidence extraction, and adapt Task 2 separately for factor identification. We also tailor aggregation to each output: we average risk-level probabilities from the 32B and 72B models, combine evidence phrases through cross-fold consensus, and calibrate factor-specific decisions through rate matching based on out-of-fold operating points. On the official leaderboard, the final system achieved a composite score of 0.7738, with 0.8089 on Task 1 and 0.6919 on Task 2. Across the evaluated configurations, three-task training performed best for Task 1a, joint training on Tasks 1a and 1b performed best for Task 1b, and task-specific training performed best for Task 2. Probability averaging further improved Task 1a when component models had complementary errors. These findings highlight the value of tailoring both training objectives and aggregation strategies to the output structure of each task within a unified language-model framework.

Legal research is a core and time-consuming legal workflow. Lawyers must identify controlling authority, verify that it remains valid, reconcile statutes and cases, and synthesize a grounded answer. Language model agents are a natural fit for this retrieval-intensive workflow, and automating even part of it would be valuable. But that value depends on reliability: a single missing authority, stale citation, or wrong legal conclusion can make an otherwise plausible answer unusable. We introduce Legal Research Bench (LRB), a benchmark of 413 open-ended U.S. legal research questions written by experts, each paired with a gold answer, supporting authorities, and a binary grading rubric. We evaluate thirteen frontier models in a harness with web search, case-law search, page parsing, and retrieval tools. We score agent responses through all-pass grading with source verification, where a response is correct only if every required criterion is satisfied and its cited authorities verify. We also validate the LLM judge against expert attorneys ensuring that benchmark scores track attorney judgment. Agents remain far from reliable: among the models we tested, the strongest, Claude Opus 4.8, is fully correct on 42.9% of questions. Performance also varies substantially by task setting: all-pass rates differ across areas of law and are lower on questions requiring reconciliation of conflicting authorities. Across models, more turns, tool calls, and inference cost do not predict higher accuracy.

Agents competing against a common opponent can often gain an edge by forming alliances, but deciding whether to do so is difficult when they lack precise knowledge about their opponent or ally. We study how uncertainty affects opportunities for alliance formation in the coalitional General Lotto game, in which two players compete aganist a shared adversary by allocating their resource budgets over separate sets of valued contests. When all agents know one another's budgets, it is known that a budget transfer from one player to the other can benefit both players. We ask whether such alliances survive when the players know the relevant budgets only up to a multiplicative factor. Under two models, one in which the players are uncertain about the adversary's budget and one in which they are uncertain about each other's budgets, we characterize exactly which games admit a transfer that benefits both players in every game consistent with what they conjecture to be possible. In both models, we show that such robust transfers exist in a set of games of positive measure for every level of uncertainty. These results suggest that alliances can remain worthwhile even when allies are largely in the dark.

Historical music restoration (HMR) has almost exclusively focused on constrained problems such as Super-Resolution or the restoration of solo pieces, under-exploring the general task of restoring orchestral historical music, which has multiple instruments. This under-exploration is largely because the HMR domain, early-20th-century recordings, has no pre-degradation ground-truth pairs, making the restoration task unsupervised and more challenging. This paper presents a supervised end-to-end orchestral HMR benchmark by exploring both the synthetic degradation functions and the end-to-end generative deep-learning restoration methods. We simulate the historical recording degradation chain more faithfully than prior work, which makes orchestral restoration into a tractable supervised problem. A latent flow-matching model trained on the resulting synthetic pairs outperforms existing HMR baselines on intrusive, non-intrusive, and subjective evaluations. We also curate and release a 9.3-hour license-free, unpaired, historical classical-music test set, along with code and audio demos.