Papers

Oda's smoothness conjecture asserts that every smooth lattice polytope has the integer decomposition property. The three-dimensional case of this conjecture was recently proven by AI. Here we give a much shorter proof of this result.

Learning symmetries through data augmentation offers an alternative to architectural equivariance, but does not guarantee that a network respects those symmetries. Measuring learned equivariance is especially difficult for intermediate representations, whose group actions are not specified. In a finite discrete setting, we show that equivariance requires representations of original and transformed inputs to determine one another exactly, yielding a test based on normalized mutual information (NMI). For high-dimensional activations, we use centered kernel alignment (CKA) as a scalable test, with an exact guarantee for orthogonal feature transformations. Across architectures and tasks, augmentation usually increases transformation consistency, showing that learned equivariance can be probed directly from representations without recovering the hidden transformation rules that relate them.

Modern mixed-integer programming (MIP) solvers repeatedly solve linear programming (LP) relaxations, but the LP state transferred from the simplex algorithm to the branch-and-bound tree is typically treated as an artifact of perturbation, presolve history, and numerical trajectory. This is harmless in pure linear programming, where alternative optimal bases certify the same bound. In MIP, however, those bases can expose distinct fractional structures, active rows, reduced costs, branching candidates, and reoptimization states. The inherited LP state is therefore an initial condition for the integer search. We exploit this initial condition to mitigate computational performance variability. The proposed framework generates alternative LP states on or near degenerate optimal faces, evaluates them within an auxiliary solver environment, and either selects a single state or executes a portfolio of candidate states in parallel. A handcrafted synthetic instance isolates the relationship between the initial LP state, the fractional profile, the branching trajectory, and the resulting node-count distribution: root-state steering reduces the weighted variability score~(WVS) by 67% over 1,000 paired seeds and produces no node-count regressions relative to default SCIP. On a separate 130-instance MIPLIB 2017 validation set with up to 50 paired seeds per instance, the top-five policy improves solve completion from 3,792 to 3,894 runs, reduces the shifted geometric mean of the recorded node count from 2,574.7 to 1,198.0, and lowers the median WVS by 9% on instances reliably solved by the baseline. These results demonstrate that the branch-and-bound phase inherits a specific LP state rather than merely an LP bound, and that the active management of this state systematically mitigates computational performance variability.

The Frobenius problem asks, given two positive, relatively prime integers $a$ and $b$, what is the maximum integer that cannot be expressed as a linear combination of $a,b$ using non-negative coefficients? This problem is resolved. However, if we add the restriction that the coefficients themselves must also be relatively prime, we create an open problem with connections to the Stern and Farey sequences and the Jacobsthal function. We prove that under this new restriction, a maximum inexpressible integer still always exists, and explore its values and properties as functions of $a,b$.

Diffusion language models are a promising and more efficient alternative to autoregressive models. However, most RL post-training techniques are not easily adapted to these models. One issue is that diffusion language models generate answers through many intermediate states, deciding at each step which masked positions to reveal and which tokens to place in them. Many existing RL methods used in this setting approximate intractable sequence likelihoods or weight every denoising step by the same advantage, without explicitly distinguishing how these intermediate states of the denoising trajectory affect the expected reward. To address this issue of trajectory blindness, we introduce PACT (Potential-Anchored Consistency of Transitions). Motivated by a reward-tilted reference distribution, PACT jointly learns the policy and a scalar potential that scores each intermediate state, anchored by terminal reward. For each sampled denoising transition, a consistency loss matches the policy-to-reference log-probability ratio of revealed tokens to the change in this potential. This formulation requires neither sequence-likelihood estimation nor importance-ratio clipping. On four tasks, PACT improves LLaDA-8B-Instruct by +9.9 points on GSM8K, +12.0 on MATH500, +55.7 on Countdown, and +72.6 on Sudoku, and reaches the best average accuracy across the four tasks reported for this model, +3.7 points over the next best method. Further compared to AGRPO, it reaches that baseline's best accuracy in about half the training compute on GSM8K.

Federated learning allows hospitals to train a shared mortality model without pooling their records, but published comparisons of aggregation strategies rarely control the evaluation protocol. This study compared seven strategies (FedAvg, FedProx, SCAFFOLD, FedAdagrad, FedAdam, FedYogi and FedCluster) on two federations built with one pipeline: 12 care units of a hospital in MIMIC-IV and 49 hospitals in eICU-CRD. Each strategy was tuned and selected on validation data by the same rule, and every comparison was repeated across 10 seeds with Friedman and Holm-corrected tests. SCAFFOLD ranked first in both federations (AUC-ROC 0.886 and 0.819) and in 17 of 18 variants, but its margin over the runner-up never exceeded 0.008, and a control sharing its local optimizer attributed most of its margin over FedAvg to drift correction. On the multi-hospital federation, federated training matched centralized training. In the care-unit federation, three protocol choices changed discrimination by more than the choice of strategy did: discharge-coded predictors (+0.038 in AUC-ROC), client-level SMOTE (up to -0.074), and evaluating the last instead of the best validation round (up to -0.084). Platt scaling on validation data repaired the risks distorted by imbalance corrections without loss of discrimination. Studies should therefore fix the protocol before ranking algorithms.

The Brout–Englert–Higgs (BEH) mechanism can be formulated in gauge-invariant relational terms without assigning physical spontaneous breaking to a local gauge redundancy. We treat the gauge orientation of the Higgs field as dynamical internal reference data and describe the gauge connection relative to this Higgs frame. On a regular patch where the frame exists, the Higgs kinetic term induces a positive-semidefinite mass-squared quadratic form on the gauge-orbit directions and reproduces the standard tree-level vector masses. A complete Higgs frame yields a gauge-invariant relational connection, whereas a partial frame is defined only up to a stabilizer group $H$ and produces Higgs stiffness only along the coset directions $G/H$. In quantum field theory the normalized frame and relational connection are nonpolynomial composite variables; physical masses are therefore characterized through suitably renormalized gauge-invariant composite correlators, or through fully dressed operators in residual-charged sectors. Applying this construction to the electroweak theory, we show that the Higgs frame relationalizes the $SU(2)_L$ redundancy while leaving a residual electromagnetic $U(1)_{\rm em}$ Gauss law: $W^\pm$ and charged fermionic modes require electromagnetic dressing, whereas neutral channels admit local gauge-invariant interpolators. This formulation connects Higgs dressings, FMS composites, and internal quantum-reference-frame language while keeping local mass generation distinct from global symmetry realization and topological obstructions.

Muon colliders are exceptionally intense sources of multi-TeV neutrinos, allowing even relatively small detectors to record large rates of neutrino interactions. In addition to serving as neutrino-ion collider for precision measurements of neutrino deep inelastic scattering, this setup allow to study even rare neutrino processes with at high statistics. In this study, we investigate the feasibility of measuring neutrino trident and on-shell W-boson production, predicting over one million trident events and one hundred thousand W-boson events in a 100 kg silicon-based active target over ten years and allowing to detect these processes for the first time. We further show that these measurements could constrain new vector and axial-vector interactions beyond projected DUNE sensitivities and provide unprecedented constraints on the high-x photon distribution through W-boson production near threshold. We also discuss strategies for signal-background discrimination and the resulting detector requirements.

This paper initiates the study of arithmetic properties of closed points of arbitrary degree on curves. We show that a pointless curve of rank $0$ can have only finitely many cyclic degree $\ell$ points for all sufficiently large primes $\ell$. More generally, we show that the residue fields of points on curves of genus at least $2$ and rank $0$ are often biased away from total ramification. The proof technique illuminates a natural intermediate conjecture between the Manin–Mumford conjecture and the Zilber–Pink conjecture for self-products of abelian varieties. This intermediate conjecture implies analogous results for other splitting profiles.

Standard semidefinite programming (SDP) solvers working in double precision can become unreliable on difficult problems, even when their output appears plausible. We study this issue for SDP problems arising from high-order moment-SOS hierarchies. Three polynomial optimization problems are analysed by comparing standard double-precision computations with reference solutions obtained using the extended-precision solver \mbox{Loraine\,64x4}. We observe two distinct patterns of numerical failure. In false convergence, the computed bounds form a smooth sequence but approach the limit at the wrong rate. In stagnation, the sequence develops a plateau or oscillations and no longer reflects the convergence of the underlying hierarchy. We introduce an empirical effective condition number based on the positive spectral components of the primal and dual solution matrices. It serves as a warning indicator rather than an error bound, and its growth accompanies numerical deterioration in our examples. We also examine the imbalance between primal and dual spectral conditioning and a Cholesky inverse residuum as potential diagnostic indicators. The results show that solver termination messages and apparently regular convergence curves may be insufficient to establish the reliability of high-order moment-SOS bounds.

To reconstruct a predictive 4D representation of the physical world from monocular video captured by a moving camera, we jointly reconstruct dynamic Gaussian radiance fields and infer their underlying physical dynamics. This paper proposes Physical SLAM, a physics-aware method for constructing 4D Gaussian fields of deformable objects and predicting their future state when subjected to external actions, given a sequence of RGB-D images. Our method separates the scene into a static background Gaussian field for camera localization and a dynamic object Gaussian field driven by a spring–mass physical model of elastic deformation, inertia, and damping. To integrate visual observations with physical dynamics, we introduce virtual observation forces that gradually pull simulated surface nodes toward observed depth points, avoiding abrupt replacement of simulated node positions. Meanwhile, the physical parameters of the spring–mass model are updated from simulation errors aggregated over the full reconstruction observation window, enabling the reconstructed dynamics to align with observations. To couple appearance, geometry, and physics, we further design a physics-aware Gaussian transformation and surface alignment strategy, where dynamic Gaussians are transformed along the deforming physical surface of the spring–mass model and constrained to remain aligned with the surface during rendering-based optimization. The resulting representation supports observed-frame rendering and action-conditioned simulation, extending SLAM maps to describe how deformable objects respond to manipulation. On the PhysTwin benchmark, our method improves reconstruction and prediction PSNR over PhysTwin by 10.4% and 11.0%, respectively. On the TUM RGB-D benchmark, it reduces average camera tracking error by 20.6%, improves SSIM by 3.6%, and reduces LPIPS by 8.3% compared with 4DGS-SLAM.

The static Love numbers of black holes vanish in classical, four-dimensional, asymptotically flat General Relativity. We study how this changes for the response of near-extremal Reissner–Nordström black holes to a minimally coupled massless scalar, at energies where the Schwarzian mode of the near-horizon throat fluctuates strongly. The black hole then responds as a quantum system with a spectrum of states. Its static response is a polarizability, given by virtual transitions between them. For $\ell=0$ the static Love number vanishes, by the shift symmetry of the scalar, but for multipoles $\ell\geq1$ the static Love numbers are nonzero: the finite rotational energy of the spin-$\ell$ intermediate states spoils the cancellation responsible for the classical result. Since the throat no longer has a classical geometry from which to read off a tidal response, we define it through the retarded Green's function, and compute its full frequency dependence at leading order in $r_+\omega$. Its conservative part requires renormalization: the near-horizon theory fixes the coefficient of its logarithm, and with it a dependence on the energy that no local term can remove, while its finite local terms are matching data of the four-dimensional theory. The dissipative part is finite and scheme independent. For $\ell\geq1$, it vanishes at low frequencies when the energy is below the rotational gap, while at the gap threshold it becomes non-analytic in frequency and produces a late-time $t^{-3/2}$ tail.

The higher Green's functions $G_{k/2}(z,w)$ on the modular curve are real-valued eigenfunctions of the hyperbolic Laplacian in each variable, with logarithmic singularities along the diagonal. Gross and Zagier conjectured that their values at CM points, after multiplication by suitable powers of the discriminants, are logarithms of absolute values of algebraic numbers. The conjecture was proved by Li, Bruinier, and Yang. Mellit proposed a complex refinement of this conjecture, predicting that the normalized CM values of a complex lift of the Green function are logarithms of algebraic numbers modulo periods. For $\Gamma=\mathrm{SL}_2(\mathbb Z)$ and $k\in{4,6,8,10,14}$, we follow Zagier's construction of CM invariants from Eichler integrals of the Eisenstein series $E_4$. By applying Maass raising operators in both variables to a normalized iterated Eichler integral of a Bergman-type kernel, we obtain a complex-valued extension $\mathcal G_{k/2}^\bullet(z,w)$ whose real part is a constant multiple of $G_{k/2}(z,w)$. We study the normalized CM values of $\mathcal G_{k/2}^\bullet$ and describe their dependence on the choice of representatives of quadratic form classes. In particular, we show that replacing the quadratic form corresponding to the first variable by an equivalent one changes the exponential of the normalized value only by a root of unity. We also prove the algebraicity of these exponentials in the special case where the second CM point is $w_*=(1+i\sqrt d)/2$, with $d>3$ and $d\equiv3\pmod4$, and the class of the quadratic form corresponding to the first variable is equal to its inverse.

Feynman's ratchet illustrates how work can be extracted from thermal fluctuations in the presence of two heat baths at different temperatures. While some experimental realizations of the ratchet effect have been studied, most lack miniaturization, which is important for some applications. Here, we propose a physical implementation of Feynman's ratchet based on Josephson junctions exhibiting the superconducting diode effect. This setup functions as a heat engine that generates work solely from a temperature gradient via thermal fluctuations. We analyze its output power and efficiency, and, through the Seebeck coefficient, identify the conditions under which the device operates in reverse, functioning instead as a Peltier cooler.

Conversational analytics lets people ask questions of records spread across many systems and get answers in plain language; it is useful only if each answer can be trusted and explained. A production system of this kind attaches a confidence score to every answer it writes over a database and uses it to show the answer, show it with a warning, or withhold it. The score's weights, caps and thresholds were set by design. This paper audits that design. On the 1,534 questions of BIRD dev, answered by GPT-5.1 with a benchmark-adapted production prompt and five samples per question, the designed score separates right from wrong answers weakly (AUROC 0.624, 95% CI 0.595-0.652) and withholds 35.7% of correct answers. Nearly all of its ranking signal comes from agreement among the samples, and most of it is lost where agreement is combined by a minimum with a checklist that drops to zero on any unsupported claim, including claims in commentary. We froze a rule in which agreement with the shown answer is the score and deterministic checks act as vetoes and caps, and pre-registered a test on Spider dev before any Spider answer existed. The primary hypothesis held: AUROC rose from 0.566 to 0.725 (paired difference +0.159, 95% CI +0.097 to +0.221), and correct answers withheld fell from 49.0% to 8.8% with PASS accuracy 95.7%. Under one matching policy the choice of agreement statistic matters little; the gain comes from how agreement enters the score. Whether added checks help depends on how correctness is defined, while they always cost unqualified coverage. A pre-registered empty-result check did not meet its effect-size criterion on Spider. Correctness here means the core answer; labels are model-judged, and a blinded two-person audit, not yet complete, will be reported in a revised version.

We studied librating orbits of stars near the Outer Lindblad Resonance (OLR) of the bar using the angle-action variables of Hamiltonian dynamics. The librating orbits change the direction of their elongation relative to the major axis of the bar in a limited range of angles. To calculate orbits, we used the model of the Galaxy with an analytical bar which reproduces well the distributions of the observed stellar velocities along the Galactocentric distance R derived from the Gaia DR3 data. We compared the values of the orbital oscillation period of a typical star obtained from the theory and with the use of numerical simulation. Both values were found to be in good agreement and are equal to 2.0 Gyr. It was shown that the change in the average orbital radius, <R>, of a star, calculated over the time interval of one radial oscillation, is regulated by the change in the direction of orbital elongation. As the orbit is tilted relative to the minor axis of the bar in the sense of Galactic rotation, the angular momentum of the star, L, decreases, on average, causing a decrease in <R>. As the orbit is tilted in the opposite direction, <L> and <R> increase. We derived a general formula for the oscillation frequency, omega, of orbits captured by libration near the OLR of the bar. It turned out that the ratio of the frequency of orbit oscillations to the epicyclic frequency, omega/kappa, is determined by the parameters q_R and <p>, where the factor q_R is the ratio of the amplitude of the radial force generated by the bar to the average radial force produced by the entire galaxy at the distance of the OLR, and the parameter <p> characterizes the average elongation of orbits captured by libration near the OLR.

The mechanism driving the evolution and dispersal of protoplanetary discs has a substantial impact on the theories of planet formation. To date, neither the viscous paradigm nor the MHD wind-driven framework have been ruled out by the observations. The statistical richness of the observed samples provides a unique opportunity to test evolutionary models through population synthesis. However, comparisons between models and observations are normally carried out by eye, and a more statistically rigorous approach has not yet been implemented. We present DEMOS, a Bayesian inference tool that aims to infer the model parameters and initial conditions that best reproduce the full distribution of global properties of disc populations. This framework relies on a comparison between observations and synthetic populations generated with the 1D disc population-synthesis code Diskpop. We tested the tool with both the MHD wind model and the viscous framework with internal photoevaporation. Applying DEMOS to synthetic populations, we find that, for both models, it accurately recovers the model parameters of disc populations and their associated uncertainties, including cases in which initial conditions are correlated with the stellar mass. In the optimal scenario in which disc masses, accretion rates and disc sizes are available, the tool can also constrain the angular-momentum transport parameter. We show that, although parameter estimates are the most precise with samples of about 100 discs, the tool also yields meaningful constraints for smaller populations of 10-30 discs. We further find that the predictions obtained with DEMOS closely align with those found by previous works. We anticipate that DEMOS will be a valuable tool to infer the initial conditions of observed disc populations, to design observational programmes, and to assess which evolutionary model best reproduces the data.

Binaries composed of a luminous OB star and a black hole (BH) are a key evolutionary link between massive stellar binaries and merging compact-object binaries, but only a handful are known and their occurrence fraction remains poorly constrained. We use multi-epoch spectroscopy from the SDSS-V Milky Way Mapper to place a first rigorous empirical upper limit on the fraction of Galactic massive ($\geq 10\,\mathrm{M}_\odot$) OB stars with a close BH companion, $f_\mathrm{BH}$. Radial-velocity (RV) variability alone cannot distinguish BH companions from the far more numerous luminous binaries. We therefore combine the SDSS-V RV variability measurements of 3130 massive OB stars with high-resolution follow-up and literature information to identify luminous companions among the RV-variable subset. Forward-modelling the detectability of OB+BH systems using BH masses and orbital parameters from binary-evolution models, we find $f_\mathrm{BH} \lesssim 3\%$ at 95% credibility. This empirical upper limit approaches, but does not exclude, theoretical predictions for the OB+BH fraction. The constraint is primarily limited by the number of RV-variable systems that can be decisively classified, highlighting the importance of complete, high-quality follow-up for future searches.

Search-based policy learning, such as Monte Carlo tree search (MCTS), offers the flexibility to trade runtime computation for data efficiency and adaptation to novel environments. For robot learning, however, applying search requires discretizing the continuous action space and associated dynamics, typically through action quantization. While reconstruction-based quantization is sufficient for representing the continuous action space alone, policy learning additionally requires the quantized actions to preserve their effects on future states and values. We show that such requirement is not necessarily captured by action reconstruction error. Instead, we prove that the error in the value function resulting from action quantization is tightly bounded by the reconstruction error of state reachability under the quantized actions. Based on these results, we propose value-preserving action quantization (VAQ), which quantizes actions by preserving state reachability. We further combine VAQ with MCTS (VAQ-MCTS) as a value-preserving, search-based policy learning framework for behavioral cloning. Using a wide range of object pushing tasks as a controlled benchmark, we evaluate VAQ-MCTS against state-of-the-art behavioral cloning methods. VAQ-MCTS matches or exceeds their performance under standard settings, while substantially improving data efficiency and robustness to heterogeneous demonstrations, while enabling zero-shot adaptation to novel environments. Hardware experiments further demonstrate the effectiveness of VAQ-MCTS, completing all 30 real-world object pushing trials with only 10 minutes of demonstration data.

Separating Alzheimer's Disease (AD), Parkinson's Disease (PD), and Healthy Controls (HC) using structural MRI remains challenging because disease signatures overlap and model performance can be strongly affected by data partitioning. We present HybridDualPathNet, a hybrid architecture combining EfficientNetV2-S for local features and Swin-Tiny for global context through a Cross-Fusion Attention (CFA) module. We evaluate the model in two experiments. First, on the Kaggle APD-3-Class dataset using an image-level split, the model achieves 98.25% accuracy. Second, on a combined Kaggle and Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset containing 17,859 slices, we use a per-source split with patient-wise separation for ADNI subjects. In this setting, the model achieves 73.37% standard accuracy and a macro AUC of 0.880. Source-stratified evaluation reveals a substantial domain gap, with 91.2% accuracy on Kaggle test images and 59.2% on ADNI test images. Temperature scaling reduces expected calibration error from 0.130 to 0.015, while split-conformal prediction achieves 89.5% coverage. Saliency analyses using Grad-CAM++, Integrated Gradients, and occlusion indicate that model activations concentrate on brain tissue. These results highlight the importance of patient-level separation, cross-dataset evaluation, and uncertainty estimation when assessing MRI classification systems. The findings also show that strong performance on a single dataset does not necessarily translate to reliable generalization across imaging cohorts.

In this paper, we study the following magnetic nonlinear Schrödinger equation \[ \left(\frac{\varepsilon}{i}\nabla-A(x)\right)^{2}u+V(x)u=f(|u|^{2})u\,\ \text{in}\,\ \mathbb{R}^{2}, \] where $\varepsilon>0$ is a parameter, $A\in C(\mathbb{R}^{2}, \mathbb{R}^{2})$ is a magnetic potential, $V\in C(\mathbb{R}^{2}, \mathbb{R})$ is a multi-well potential, and $f\in C^1(\mathbb{R}, \mathbb{R})$ exhibits exponential critical growth. By variational methods and a penalization technique, we prove the existence and multiplicity of nontrivial solutions for sufficiently small $\varepsilon>0$. If, in addition, $A\in C^1(\mathbb{R}^2,\mathbb{R}^2)$, we show that these solutions have multiple peaks and we establish their concentration behavior and exponential decay.

Differentially private (DP) pipelines often combine heterogeneous mechanisms, yet their guarantees are best analyzed using different notions of privacy (e.g., bounded range, zCDP, ADP). When composing them, we have to analyze the guarantees through a reduced common representation that loses accounting precision. This issue is particularly relevant in DP synthetic data generation, where data preprocessing, generation algorithms, and DP-SGD could appear within the same pipeline. We develop novel tools for tight end-to-end accounting of such compositions using privacy loss distributions (PLD), $f$-DP, and Gaussian DP (GDP). In particular, we derive PLD and GDP analyses of bounded-range mechanisms such as the Exponential mechanism. Building on these theoretical results, we introduce AIM-GDP, a GDP-native variant of AIM (a state-of-the-art DP synthetic data mechanism) that enables easy composability with complex pipelines via PLD, attack-aware risk calibration, and tight end-to-end auditing while consistently achieving better utility; and MixCal, an algorithm for calibrating adaptive mixed compositions using PLD accounting. Overall, our tools enable tighter end-to-end accounting and thus less noise at the same privacy level in DP synthetic data pipelines.

This study extends the optical/X-ray correlation from black hole low-mass X-ray binaries (XRBs) to active galactic nuclei (AGN), bridging scales from stellar-mass to supermassive black holes (SMBHs). In XRBs, optical emission predominantly originates from the thermal reprocessing of high-energy photons on the surface of the geometrically thin, optically thick accretion disk. The reprocessed optical luminosity follows a power-law relation of the form $L_{V} \propto L_{X}^{\beta} a$, with $\beta \sim 0.5$, where $L_X$ is the X-ray luminosity and the orbital separation $a$ serves as a proxy for the disk size. To test the universality of this relationship across all mass scales, we analyzed quasi-simultaneous X-ray and optical flux data from a sample of XRBs and AGN that have estimates of the radius of the optically emitting disk. To bridge the structural differences between these populations, we utilized the characteristic optical/X-ray time lag derived from reverberation mapping as a physical proxy for the reprocessing radius. Our analysis reveals a tight, global correlation across all black hole masses, described by $\log L_V = 0.48\log L_X + 0.96\log a_R + 6.5$ with a scatter of $\sim 0.20$ dex, where $a_R$ is the reprocessing radius derived from the lag. Incorporating a geometric correction for the disk inclination angle ($\cos i$) slightly reduces this scatter to $0.18$ dex. These results suggest that X-ray reprocessing is scale-invariant across accretion disks, from stellar-mass black holes to SMBHs. This scaling relation provides new constraints for models of X-ray emission regions in accreting black holes.

We report the discovery and timing of PSR J2147-3604, a fully recycled millisecond pulsar with a spin period of 3.2 ms, discovered in a reprocessing of the 350-MHz Green Bank North Celestial Cap (GBNCC) survey and later serendipitously confirmed by the MeerKAT TRAPUM UHF survey of Fermi-LAT unidentified sources. We followed up this pulsar with Murriyang/Parkes radio telescope and obtained a phase-connected timing solution spanning one year, revealing a 4.5-day binary orbit with very low eccentricity (e ~ 1.9 x $10^{-5}$). A follow-up search in LAT data revealed gamma-ray pulsations, allowing for a joint radio and gamma-ray timing analysis. Assuming a neutron-star mass of 1.5 Msun, the inferred median companion mass (~0.2 Msun) is consistent with a helium white dwarf. The pulsar has a low dispersion measure of 17.0 pc cm$^{-3}$ and shows strong scintillation, leading to it being detected in ~72% of the observations covering the 704-1174 MHz band with Murriyang. The pulsar has a narrow duty cycle of ~6.5%, which leads to individual time of arrival (ToA) uncertainties of ~4.5 us for a 10 min integration with MeerKAT, suggesting this pulsar is well suited for inclusion in PTA experiments.

Generative discovery in science requires coupling high-level scientific reasoning with the ability to generate samples respecting the structure of a design space. Yet these capabilities remain largely separated: foundation LLMs encode broad chemical and biological knowledge, while domain-specific flow- and diffusion-based generators capture the priors needed to produce valid sequences and geometries. We study how to align such scientific generators to high-level natural-language discovery goals without requiring an explicit proxy reward, which is often unavailable or poorly captured by learned predictors in experimental sciences. We introduce LLM-to-Flow Distillation (LFD), where a frozen LLM acts as an adaptive teacher for a domain-specific generator. Given a goal such as 'generate peptides with promising anticancer activity', the teacher constructs a curriculum of intermediate sub-goals aware of the flow's current capabilities, and progressively directs them toward the original discovery goal. At each round, the LLM judges sampled candidates under the current objective. These judgments induce preferences that are iteratively distilled into the generator through DPO. This procedure provides dense learning signals even when the initial model places negligible probability on designs satisfying the final goal, enabling adaptation toward poorly covered, out-of-distribution regions. Experiments on therapeutic peptides and small-molecule structure design demonstrate that LFD enables a new form of generative control: aligning arbitrary domain-specific scientific generators to scientific discovery tasks expressed via high-level natural-language goals.

This paper develops a Willis-type thermal homogenization framework for two-dimensional particulate composites under harmonic heat transfer and establishes a unique effective constitutive relation without introducing residual fields. The local thermal fields are obtained using the dual-reciprocity inclusion-based boundary element method, in which two eigen-fields are introduced to simulate the mismatches in thermal conductivity and heat capacity. Instead of retrieving effective kernels from several independent macroscopic solutions generated by residual fields, the complete constitutive operator is constructed upon relations between eigen-fields and generalized macroscopic responses. The effective constitutive relations are validated against a one-dimensional analytical solution and an extended Fourier-series homogenization scheme. The Fourier formulae further establish the opposite-wavenumber reciprocity relations between two coupling tensors and reflection-symmetry parity relations, which explain when transverse coupling components vanish and the sign reversal of off-diagonal conductivity components under reflection. Numerical results further demonstrate that the de-phased average is required to extract physically consistent effective thermal properties under the Bloch conditions, and reveal their nonlocal frequency/wavenumber-dependent behavior. The framework provides a general computational method for Willis-type homogenization of particulate composites and is extendable to three-dimensional, multiphysical problems.

Fast, high-fidelity molecular generation requires accurate chemical choices and efficient sampling. Chemical building blocks reduce graph size, but make each node a choice among tens of thousands of related identities. We introduce MIRRA (Molecular Identity Representation, Refresh and Assembly), a discrete flow-matching generator that learns from chemical knowledge and evaluates identity probabilities only when needed, while keeping every identity in its vocabulary available. Chemical features share information across related blocks, and learned identity corrections distinguish them in molecular context. By deciding which positions will refresh before computing their probabilities, the sampler avoids unused predictions while preserving each sampling transition. Against the strongest baselines in each task, MIRRA reduces chemical and graph distribution errors on ZINC250k by 29-31% and 53-60%, respectively, with generation up to 8.6 times faster. Property targeting yields an estimated 7.8-20.8 times more distinct novel hits per second. On a held-out polymer test set, MIRRA achieves 100% periodic validity with known chain topology, versus 57.0% for a polymer variational autoencoder. These results demonstrate fast, high-fidelity generation with the full chemical vocabulary available throughout sampling.

The original EneMonIE proposal describes a Web-based semantic system for continuously monitoring energy trends through targeted information extraction from heterogeneous online media. This article presents EneMonIE-Backend, an implementation-derived redesign that converts that conceptual pipeline into a modular, executable, and application-independent backend. The architecture separates five concerns: data sources, acquisition and normalization, semantic processing, knowledge organization, and persistent storage. It operationalizes configurable connectors for RSS feeds, URL collections, arXiv, social-media JSON files, and open-data CSV files; it also introduces question-driven Web search and scraping. Relevant documents pass through energy categorization, named entity recognition, temporal and event extraction, source sensitive sentiment analysis, and extractive summarization. The resulting semantic representations support entity co-occurrence graphs, cross-document interlinking, cluster-level information fusion, and relational persistence in SQLite. Three grounded question-answering paths are exposed at the backend level: live Web evidence, the stored corpus, and selected configured resources. Answer generation uses a language model when configured and otherwise falls back to deterministic extractive output. Unlike the original proposal, the presented architecture deliberately excludes the user-interface and server-side application layers, allowing the semantic core to be embedded in different applications. A deterministic smoke validation confirms relevance filtering, persistence, clustering, evidence retrieval, and document deletion. The work provides a reproducible architectural basis for future evaluation on larger multilingual energy corpora.

Can a vision-language model tell left from right? Modern VLMs solve visual puzzles and follow hour-long videos, yet when asked whether the camera moved left or right, most do no better than a coin flip. Adding camera-motion data or geometric representations has been tried, separately, and neither closes the gap. We argue that camera-motion understanding needs two conditions at once: the representation must couple off-screen camera motion with on-screen flow dynamics, and it must be interpreted by the language model rather than trained as a standalone pose or depth estimator. We propose CameraVLM, which pairs a motion-state representation with state-language alignment. Camera tokens capture how the camera moves off screen, flow tokens capture how the subject and background move on screen, and a two-stage training strategy keeps language in the loop so that both are learned together with camera-motion semantics. With only 9B parameters, CameraVLM outperforms all open-source and camera-motion-specific models by a wide margin across three benchmarks and more than 20 baselines, and matches or surpasses GPT-6 Astra as well as human viewers in a 48-participant study. Project page: https://3dagentworld.github.io/CameraVLM/

Sudden beam loss (SBL) events at the SuperKEKB collider result in the loss of a large fraction of the stored beam within just a few revolutions, on a timescale comparable to or shorter than the 20-30 {\mu}s latency of the existing abort system. Such events have damaged collimators and threaten Belle II subdetectors. Bunch Oscillation Recorders (BORs) capture transverse bunch positions at each revolution, which often show sizable oscillations that precede significant beam current loss. We test whether these precursors can be detected earlier than a charge-based reference that mimics the existing abort sensors. We apply several oscillation-based anomaly detection algorithms to 208 SBL events recorded in the low-energy ring during the 2024 run period, ranging from simple rolling statistics to machine learning models trained to forecast bunch oscillation. On 173 events not used for training or calibration, the best-balanced model (DeepLSTM) triggers before the charge-based reference in 86-90% of events across five training seeds, with a mean lead time of 25-30 {\mu}s. The only trigger classified as a false alarm fires on an oscillation that begins about 60 turns before the abort and persists until the beam is lost, making it more likely an early precursor than a spurious trigger. These results establish bunch oscillations as a viable early abort trigger and motivate a shadow-mode deployment to measure the false-alarm rate in continuous operation.