AI for Science

The ability to compute electron-phonon coupling from first principles, using density functional perturbation theory and interpolation techniques, has enabled predictive calculations of electronic transport coefficients in crystalline materials. However, these methods are still computationally expensive. Here we present two inexpensive methods to obtain thermoelectric transport properties of semiconductors using a small number of electron-phonon matrix elements calculated from first principles. The first method combines models for coupling of electrons with different phonon modes whose parameters are obtained from $\sim 10$ matrix elements per electronic band and phonon mode calculated from first principles. Within this method, we formulate the acoustic deformation potential model for arbitrary crystal symmetries and band extrema locations. The second method uses machine learning to interpolate $\sim 100$ electron-phonon matrix elements per electronic band and phonon mode on dense reciprocal space grids in the parts of the Brillouin zone relevant for transport. We apply both methods to two-dimensional MoS$_2$ and show very good agreement with the state-of-the-art method. The calculated thermoelectric properties also agree well with experiments. We find that the machine-learning method is more accurate and straightforward to implement compared to the model approach.

In this study, the structural, mechanical, electronic, optical, and thermoelectric properties of the novel lead-free halide double perovskite series beta2SnGeX6 (beta = K, Rb; X = Cl, Br, I) are systematically investigated using density functional theory (DFT). Calculated formation energies, Tolerance factors, and octahedral factors confirm that all six compounds exhibit robust thermodynamic stability within a highly symmetric cubic geometry. Mechanical analysis derived from elastic parameters characterizes the entire series as fundamentally ductile, ensuring high processing elasticity and resistance to micro-cracking during device manufacturing. Electronic band structures reveal direct bandgaps showing exceptional composition-dependent tunability from 1.44 eV down to 0.64 eV via progressive halogen substitution. The wide gap chloride variations are optimized for single-junction photovoltaic absorbers, while the narrower-gap bromide and iodide analogs show immense promise for tandem solar architectures and near-infrared photodetectors. Thermoelectrically, heavy constituent atoms introduce strong lattice anharmonicity and intense high-temperature Umklapp phonon scattering, significantly suppressing lattice thermal conductivity. Combined with low carrier effective masses that optimize electrical transport, the iodide compounds achieve higher power factors and outstanding dimensionless figures of merit (ZT = 2.4 for K2SnGeI6 at 1000 K). Ultimately, these lead-free double perovskite family emerges as an environmentally benign and versatile platform for next-generation green optoelectronics and solid-state waste-heat recovery.

Iron germanium telluride (FGT; FemGenTe2) compounds have attracted significant interest due to their layered van der Waals structure, relatively high Curie temperature, and tunable magnetic properties. Chemical vapor deposition (CVD) is a particularly promising synthesis route owing to its simplicity, low cost, potential for scalability, and widespread adoption in the semiconductor industry, yet it has not been used previously to synthesize FGT with dopants. Here, we report CVD synthesis of both undoped and Ni-doped FGT nanosheets on SiO2/Si substrates. By adjusting precursor molar ratios, we synthesized Ni-doped FGT with multiple Fe concentrations and a 4% Ni-to-Fe ratio. X-ray photoelectron spectroscopy depth profiling further demonstrates that Ni is present in the bulk of the crystals. This straightforward, low-cost, and CMOS-compatible approach demonstrates a route to Ni-doped FGT nanosheets, establishing a foundation for future characterization of Ni-doped FGT and its potential integration into spintronic devices.

Proton dissociation constants (pKa) are critical for functional molecule discovery and molecular modeling. Building on iBonD, the largest experimental pKa database established, we and other researchers have developed several methods including machine-learning-based empirical prediction and high-accuracy energy calculations. Despite this foundation, the rapid augmentation of high-quality pKa data remains fundamentally constrained. As part of this work, we performed large-scale regression-based pKa prediction on unlabeled molecular datasets using a collection of extensively optimized machine-learning models. The results indicate that, since the feature distributions of unlabeled molecular datasets, the pKa data distribution approximates normality, with extreme scarcity of tail-region samples. Although such augmentation is highly valuable for improving overall data availability and predictive modeling, it remains insufficient for efficiently discovering molecules with broad-spectrum pKa properties. To address this, we explore the targeted generation of molecules with sparse pKa properties from the vast chemical space. Given that traditional continuous latent space VAE-RNN methods for molecular generation suffer from insufficient stability and fail to demonstrate clear advantages in complementing sparse data, we design and implement a quantum-assisted sparse-pKa molecular generation. Feasibility is validated on a simulated quantum annealer, and superior extreme-value sampling is further achieved on physical coherent Ising machines (CIMs). (to be continued)

The GMTKN55 data set is a collection of standard benchmarks used in molecular quantum chemistry that spans small- and large-molecule thermochemistry, reaction barriers, and non-covalent interactions. Herein, we identify a flaw in the weighted mean absolute deviation (WTMAD) definitions commonly used to quantify performance of various electronic-structure methods for the GMTKN55 set, which under-weight some of its component benchmarks by orders of magnitude. A new WTMAD-4 metric is proposed, based on typical errors observed for a set of ten minimally empirical dispersion-corrected density-functional approximations (DFAs), ensuring fair treatment across all benchmarks. The performance of 115 DFAs is then reassessed using WTMAD-4 and we highlight a literature example where a DFA parametrised by minimising WTMAD-2 underperforms for benchmarks marginalised by that metric.

High-quality InAs quantum wells grown on InP are a promising platform for topological quantum information processing due to their large g-factor, strong Rashba spin-orbit interaction, and their compatibility with in-situ-deposited superconductors. In this work, we investigate InAs/InGaAs quantum wells grown on InP (001) wafers, focusing on how the layer structure and strain influence the electronic properties and surface morphology. By combining quantum transport measurements with atomic force microscopy, we show that the layer design predominantly affects the mobility anisotropy, which aligns well with the surface morphology. Surface characterization further reveals the mechanism of quantum well collapse when the layer thickness exceeds the strain limit. In addition, transport measurements demonstrate that quantum confinement has a clear impact on band nonparabolicity.

The relativistic spin-momentum locking has been proven in time-reversal-breaking classes of materials with zero net magnetization in the non-relativistic limit, such as altermagnets and other non-collinear magnets. Using density functional theory calculations, we aim to show relativistic spin-momentum locking in ferromagnets, focusing on a broad class of ferromagnetic materials with magnetic sites connected by rotational symmetry, and compare with fcc Ni. In SrRuO3, the antisymmetric exchange interaction produces a spin canting orthogonal to the easy axis, while in all other cases, spin canting is forbidden. Even when the canted magnetic moment in real space is forbidden, relativistic spin-momentum locking shows sizable contributions in k-space. Using prototypical ferromagnets such as orthorhombic SrRuO3, hexagonal CrTe and CrAs with the NiAs crystal structure, half-Heusler MnPtSb, and fcc Ni, we demonstrate that relativistic spin-momentum locking can generate strong effects in ferromagnets. Subdominant components of centrosymmetric ferro-magnetic materials with magnetic sites connected by rotational symmetry host spin-momentum locking similar to altermagnets, while noncentrosymmetric MnPtSb hosts relativistic p-wave due to the spin-orbit coupling. Fcc Ni shows a more complex behavior with a combination of two spin-momentum locking patterns characteristic of altermagnets. Because ferromagnets typically have larger bandwidths than altermagnets, they provide a promising platform for observing even-wave relativistic spin-momentum locking and associated emergent phenomena. From an application standpoint, relativistic spin-momentum locking governs symmetry-allowed spin Hall currents, spin photocurrents, and other momentum-dependent spin responses in k-space.

We report on a study of the interacting phase diagram of $3.65^\circ$-twisted WSe$_2$ at moiré hole filling $ν=1$, in which we find previously-overlooked types of magnetism. Specifically, in part of the phase diagram we obtain a magnetic order parameter which modulates in space with four different non-zero wave vectors, corresponding to the three $M$-points and one $K$-point of the moiré Brillouin zone. These multi-Q orders, which can be coplanar or non-coplanar, are continuous deformations of the $120^\circ$ spin-valley anti-ferromagnet (AFM), where the unit cell has expanded by a factor of four. Interestingly, we find that the multi-Q states are stabilized for experimentally relevant values of interaction strength and displacement field, and are accompanied by a softening of the spin fluctuations near the $M$-points of the moiré

Altermagnets exhibit momentum-dependent spin splitting without net magnetization, combining characteristics of both ferromagnets and antiferromagnets, making them highly interesting for spintronics applications. CrSb is a prime candidate with a high Néel temperature ($\sim700$~K) and a large exchange-driven splitting of $\sim0.6$--1~eV. Using ab-initio calculations, we consider slabs of various orientations in the ultrathin limit. We find that (100) oriented slabs have spin-degenerate bands. In (0001) oriented slabs, the exchange-driven altermagnetic spin splitting collapses, but including spin-orbit coupling restores a residual anisotropic splitting of $\sim70$~meV. In contrast, the (110) oriented slabs show an altermagnetic spin splitting of $\sim400$~meV, and emerges as a robust candidate for realizing large, exchange-driven altermagnetism

Characterizing electronic band structures requires precise knowledge of wave functions and their quantum geometry. Here, we introduce a tomography method to reconstruct the density matrix of electron states from quasiparticle interference maps around single impurities. We consider two-orbital models on a honeycomb lattice, relevant to graphene heterostructures and direct-gap semiconductors. For on-site impurities, backscattering between time-reversed states directly maps the density matrix populations and coherences into distinct orbital contributions in the interference map. While local probes usually lack orbital resolution, these orbital contributions transform under distinct symmetry group representations and can thus be disentangled to reveal the density matrix and quantum geometric tensor of the scattering states. This establishes impurities as tomographic probes for band structures in scanning tunneling microscopy using conventional, unpolarized tips.

We demonstrate that all anti-self-dual Einstein metrics with non--zero cosmological constant $Λ$ locally arise from solutions of a single second order PDE introduced by Lipstein and Nagy. We show how this equation fits into the hyper--heavenly formalism of Plebański, and establish a Lax pair. Finally we show how Plebański's second heavenly equation arises in the limit as $Λ\rightarrow 0$.

We analyze a two-dimensional Bose-Fermi mixture at zero temperature in the presence of a tunable Bose-Fermi attraction. We adopt a diagrammatic T-matrix approach and study the behavior of several thermodynamic quantities for the two species as functions of density, mass ratio, and coupling strength. These include the chemical potentials, the boson momentum distribution function, the condensate density, and Tan's contact parameter. We analytically demonstrate that the present T-matrix formalism recovers the correct second-order perturbative expansion of the chemical potentials in the weak-coupling regime, and test it numerically. The near-universal behavior of the condensate fraction already found in prior work for the mass-balanced case is confirmed for different masses and becomes even more accurate when the boson mass is large. The mass imbalance emerges as an additional control parameter that qualitatively affects the bosonic momentum distribution. In particular, we found that it can be used to allow for the experimental observation of a peculiar peak in the boson momentum distribution at finite momentum.

Adaptive Hamiltonian learning is central to calibrating and characterizing quantum devices. In an adaptive controller, choosing the next experiment is itself a computation. Bayesian design rules are recomputed after every posterior update, and that step can take seconds. Across hundreds of shots, those seconds become a significant wall-clock cost for adaptivity. We introduce SymQNet, an amortized reinforcement-learning approach for low-latency adaptive Hamiltonian learning. SymQNet learns a posterior-conditioned acquisition policy offline, then uses a fast policy forward pass online while retaining Bayesian posterior feedback. On transverse-field Ising benchmarks, SymQNet substantially reduces acquisition latency relative to bounded Fisher-information search and bounded two-step Bayesian active learning by disagreement (BALD). At five qubits, it reduces acquisition-only decision latency by $47.1\times$ and $72.6\times$ relative to these online baselines; at twelve qubits, full simulated steps take $1.02$ s for SymQNet versus $13.27$ s for bounded two-step BALD. Overall, we show that learned acquisition can make adaptive Hamiltonian learning practical for repeated low-latency workloads.

Quantum circuit routing is a key step in compiling programs for noisy intermediate-scale quantum processors. Routes that appear efficient by standard overhead metrics can still lose fidelity when they pass through poorly calibrated couplers. We study a calibration-aware graph reinforcement-learning router that uses same-day IBM Heron r2 calibration data to choose hardware-edge SWAPs. We train the policy with proximal policy optimization and evaluate it with exact simulated fidelity across nine Munich Quantum Toolkit (MQT) Bench circuits and three calibration snapshots. Across these evaluations, pooled mean exact fidelity is $0.727$, compared with $0.440$ for SABRE-best20 and $0.481$ for target-aware SABRE. We observed that fidelity gains came with higher routed two-qubit counts and were concentrated in 5 qubit and 8 qubit circuit families; under the fixed tree action graph, all 10 qubit families favored SABRE-best20. Overall, our results show that calibration-aware learned routing can improve fidelity beyond gate-count-driven compilation.

We develop a data-driven framework for estimating optimal linear response of nonlinear dynamical systems. Our approach is based on kernel-smoothed approximations of the transfer/Koopman operators of the system, built from possibly high-dimensional observations along trajectories. Combining these operator approximations with the theory developed in [Antown et al. (2018), J. Stat. Phys., 170(6), 1051-1087], we formulate a computationally tractable optimization problem for the optimal infinitesimal perturbation realising any desired manipulation of the spectrum. We also introduce a notion of optimal-response vector fields for visualising, and physically interpreting, the system's response to the optimal perturbation under arbitrary observations. Our focus is on finding perturbations that optimally increase the frequency or optimally suppress the decay of correlations of almost-cycles or almost-invariant sets associated with the eigenvalues of the kernel-smoothed transfer operator. We illustrate our approach with applications to low-dimensional periodic and chaotic systems, as well as a high-dimensional example involving the El Nino Southern Oscillation in a comprehensive Earth system model. In these examples our approach discovers nontrivial optimal perturbations of the system, which are post hoc natural and consistent with the desired dynamical objectives.

Determining the neutrino mass ordering remains a central open problem in particle physics. While next-generation long-baseline experiments are expected to resolve this question, current data provide limited sensitivity because the spectral differences between normal and inverted ordering are subtle and entangled with parameter degeneracies. We investigate a machine-learning strategy for mass-ordering determination using a feed-forward neural-network classifier trained on synthetic long-baseline datasets generated with three-flavour oscillation probabilities, matter effects, and statistical fluctuations. We evaluate the classifier against standard $χ^2$ and $\log\mathcal{L}$ approaches using common discrimination metrics, including receiver-operating-characteristic curves, to quantify sensitivity and to illustrate how operating points can be selected to prioritise purity or efficiency. We find that the neural network achieves performance comparable to conventional fits for the scenarios studied, providing a flexible, independent cross-check of established analyses. The framework can be extended to incorporate systematic uncertainties and to explore joint inference of oscillation parameters, and it may also serve as a pedagogical tool for introducing machine-learning methods in neutrino physics.

The realization of Hawking radiation in optical analogs has historically focused on kinematic observables, such as the effective temperature determined by the horizon's surface gravity. A complete thermodynamic description, however, necessitates a rigorous definition of entropy and irreversibility, which has remained elusive in Hamiltonian optical systems. In this work, we bridge this gap by introducing an operational entropy for solitonic event horizons, derived from the spectral partitioning of the optical field into coherent solitonic and incoherent radiative subsystems. The emission of resonant radiation, driven by the breaking of soliton integrability under higher-order dispersion, is the fundamental mechanism for entropy production. Numerical simulations of the generalized nonlinear Schrödinger equation (GNLSE) demonstrate that, in a coarse-grained sense, this process obeys a generalized second law (GSL), $ΔS_{\mathrm{tot}} \ge 0$, robustly across a wide range of soliton orders and dispersion strengths. These results show that event horizons in dispersive field theories behave as consistent nonequilibrium thermodynamic systems, and that the relevant entropy is accessible from laboratory spectral measurements.

We analyse a generalisation of Zöttl and Stark's model of active spherical particles [Phys. Rev. Lett. 108, 218104 (2012)] and prolate spheroidal particles [Eur. Phys. J. E 36(1), 4 (2013)] suspended in cylindrical Poiseuille flow, to particle dynamics in an arbitrary unidirectional steady laminar flow through a straight duct geometry. Our primary contribution is to describe a Hamiltonian formulation of these systems and provide explicit forms of the constants of motions in terms of the arbitrary fluid velocity field. The Hamiltonian formulation provides a convenient and robust approach to the computation of particle orbits whilst also providing new insights into the dynamics, specifically the way in which orbits are trapped within basins defined by a potential well. In addition to considering spherical and prolate spheroidal particles, we also illustrate that the model can be adapted to oblate spheroidal particles.

High-fidelity simulation of spatiotemporal dynamics is computationally prohibitive, necessitating efficient super-resolution techniques to reconstruct high-resolution data from coarse-grained inputs. Traditional data-driven methods often lack physical constraints, and simple physics-informed learning struggles with irregular spatial geometries and intricately evolving temporal dynamics. To tackle these challenges, we propose a Physics-augmented Koopman-enhanced Graph Convolutional Network (P-K-GCN) for spatiotemporal super-resolution on irregular geometries. Specifically, a continuous spline-based GCN is first designed to extract spatial dependencies directly from coarse graph, and Koopman operator theory is incorporated to project the nonlinear dynamics into a compact latent space where temporal progression is linearized. Second, we augment the optimization objective with a physics-based loss to force the data-driven reconstructions to adhere to physical laws for improving predictive fidelity and robustness. Finally, we provide a rigorous theoretical analysis, establishing that the physics augmentation and Koopman regularization mathematically guarantees a reduction in super-resolution error by diminishing Rademacher complexity and tightening generalization bounds. We evaluate our framework on reconstructing spatially high-resolution cardiac electrodynamics across a 3D heart geometry from sparse low-resolution measurements. Numerical experiments demonstrate that our method achieves superior accuracy compared to baseline models.

The quality of recorded data depends on the stability of the sensor system that acquires it. Sensor motion and aging can degrade the performance and stability of downstream data-driven methods. We present a Wasserstein-GAN-inspired approach for unsupervised inference of physically interpretable transformation parameters that map a changed detector response distribution back to a nominal reference distribution. In contrast to standard generative modeling, the generator is used as a learnable calibration transformation whose trainable weights represent the sought parameters, while the critic provides a distributional distance signal via the Wasserstein objective. We validate the approach on a tracking-detector toy model with controlled layer shifts and demonstrate its application on high-granularity Geant4-simulated calorimeter data with cell-wise aging effects. The method recovers aging coefficients for individual cells with correlation to ground truth and improves agreement between calibrated and reference energy-sum distributions, while exhibiting the expected degradation at increasing channel-to-channel noise levels. These results indicate that adversarial distribution matching can serve as a data-driven component of calibration strategies in settings where direct labels for degradation parameters are unavailable.

The Remeika series superconductor, $(\text{Ca}_x\text{Sr}_{1-x})_3\text{Rh}_4\text{Sn}_{13}$, shows a rare nonmagnetic quantum critical point (QCP) associated with the continuous charge-density wave (CDW) and structural transition under the ``dome'' of superconductivity achieved by tuning composition and applying pressure. Here we use a nonmagnetic point-like disorder induced by 2.5 MeV electron irradiation to suppress the CDW and drive the system to and even beyond the QCP. This conclusion is based on a clear evolution of temperature-dependent resistivity, $ρ\left(T\right)$, from the Fermi liquid to the non-Fermi liquid regime with increasing amount of disorder. Starting on the CDW side, below the suggested QCP concentration of $x_c=0.9$, added disorder resulted in a progressively larger linear term and a reduced quadratic term in $ρ\left(T\right)$. Nearly perfect $T-$linear dependence is observed at the dose at which long-range CDW order is suppressed to $T=$0, consistent with the expectations. We refine the QCP location in this system and place it in the interval between $x=$0.75 and 0.85. Our results strongly support the concept that the disorder can tune the system to the quantum critical regime and even beyond. It follows from the argument by Imry and Ma that any ordered phase is unstable toward quenched disorder. Introduced in a controlled way, this disorder becomes a novel non-thermal tuning parameter likely applicable to a variety of different systems.

Disorder in magnetic systems typically suppresses long-range order, promoting short-range states such as spin glasses and magnetic clusters. This is particularly prominent in high-entropy materials, characterized by the random distributions of local magnetic entities and exchange interactions. However, in rare exceptions, long-range magnetic order can persist in high-entropy systems, while the microscopic characters and underlying mechanisms remain elusive, especially the magnetic behaviors of individual elements. Here, combining neutron diffraction and resonant soft x-ray scattering, we have conducted an element-specific investigation into the magnetic order of a high-entropy honeycomb-lattice van der Waals material (Mn1/4Fe1/4Co1/4Ni1/4)PS3. Despite significant atomic disorder, long-range zigzag antiferromagnetic order is observed below 72 K, with all four transition-metal elements participating in a unified phase transition. However, the spin orientations of various elements are distinct, attributed to the competition between single-ion anisotropies and exchange interactions. Our findings showcase a novel form of long-range magnetic order with disordered spin orientations, which is synergically stabilized by distinct magnetic elements in a high entropy magnet, offering a new paradigm for understanding complex magnetic systems.

Generating quantum data by learning the underlying quantum distribution poses challenges in both theoretical and practical scenarios, yet it is a critical task for understanding quantum systems. A fundamental question in quantum machine learning (QML) is the universality of approximation: whether a parameterized QML model can approximate any quantum distribution. We address this question by proving a universality theorem for the Many-body Projected Ensemble (MPE) framework, a method for quantum state design that uses a single many-body wave function to prepare random states. This demonstrates that MPE can approximate any distribution of pure states within a 1-Wasserstein distance error. This theorem provides a rigorous guarantee of universal expressivity, addressing key theoretical gaps in QML. For practicality, we propose an Incremental MPE variant with layer-wise training to improve the trainability. Numerical experiments on clustered quantum states and quantum chemistry datasets validate MPE's efficacy in learning complex quantum data distributions.

In this paper we prove a lower bound result for extremely large values of $L(\frac{1}{2},χ_p)$ with prime numbers $p\equiv 1\pmod 8$.

We study almost complete intersection ideals in a polynomial ring, generated by powers of all the variables together with a power of their sum. Our main result is an explicit description of the reduced Gröbner bases for these ideals under any term order. Our approach is primarily combinatorial, focusing on the structure of the initial ideal. We associate a lattice path to each monomial in the vector space basis of an Artinian monomial complete intersection and introduce a reflection operation on these paths, which enables a key counting argument. As a consequence, we provide a new proof that Artinian monomial complete intersections possess the strong Lefschetz property over fields of characteristic zero. Our results also offer new insights into the longstanding problem of classifying the weak Lefschetz property for such intersections in characteristic $p$. Furthermore, we show that the number of Gröbner basis elements in each degree is connected to several well-known sequences, including the (generalized) Catalan, Motzkin, and Riordan numbers, and connect these numbers to the study of entanglement detection in spin systems within quantum physics.

We consider an inverse boundary value problem for the biharmonic operator with the first order perturbation in a bounded domain of dimension three or higher. Assuming that the first and the zeroth order perturbations are known in a neighborhood of the boundary, we establish log-type stability estimates for these perturbations from a partial Dirichlet-to-Neumann map. Specifically, measurements are taken only on an arbitrarily small open subsets of the boundary.

Transversally Kähler foliations are a generalisation of Kähler manifolds, appearing naturally in the complex non-Kähler setting. We give a self-contained proof of how the classical methods used in the proof of the Aubin-Yau theorem adapt to the transversally Kähler case under the homological orientability condition. We apply this result to obtain a new, simpler proof of the already known Vaisman Aubin-Yau theorem.

On biomedical tabular data, flexible models such as deep networks, gradient-boosted trees, and kernel methods are repeatedly matched or beaten by linear and logistic regression given the same features. The usual reaction is to treat this as a model-side shortfall, to be fixed with more data, a better architecture, or tuning, on the assumption that the nonlinear structure is there and the model has failed to capture it. We argue that these fixes cannot help when the binding limit is the measurement rather than the model, as it frequently is in biomedicine. Additive noise blurs the population-optimal predictor, and because blurring removes a function's fine, rapidly varying detail before its broad shape, it erases nonlinear structure faster than linear structure. A degree-$k$ interaction is attenuated by the $k$-th power of feature reliability, while the linear part is attenuated only once. At the reliabilities typical of biomedical measurement, the nonlinear advantage can vanish even when the underlying biology is strongly nonlinear, and what the noise removes cannot be recovered by a larger cohort or a more flexible model, only by better measurement. The nonlinearity is hidden, not absent, and a tie between linear and flexible models is not by itself a verdict on the biology. These pieces are classical, drawn from measurement-error statistics, psychometrics, and Gaussian analysis, and we assemble them into an exact excess-risk identity. Measurement reliability is one of three conditions, alongside sample size and feature representation, that must align for a flexible model to help, and together they leave only a narrow window that most biomedical tasks fall outside. Across 140 UK Biobank tasks, the gap between flexible and linear models, where it exists, carries the predicted noise signature, and the three conditions can be separated by intervention but not by a benchmark alone.

Motivation: Noisy labels are a common challenge in molecular property prediction because molecular annotations are often obtained from assays, curated databases, or weak annotation pipelines rather than directly observed clean biological states. Treating recorded labels as reliable supervision can cause models to memorize corrupted observations and learn misleading molecular evidence. In multimodal molecular representation learning, this issue can be amplified by graph-text fusion or alignment, which may propagate label-induced errors across modalities. Results: We propose MOLAR, a noise-aware framework for learning multimodal molecular representations from noisy labels. MOLAR separates latent clean-property inference from recorded-label observation: graph and text views contribute residual evidence to a clean-property distribution, and a categorical label-observation channel maps this distribution to recorded labels for training. This formulation derives posterior label reliability and modality-specific molecular evidence from the model. Experiments on naturally noisy molecular benchmarks and controlled label-flipping benchmarks show that MOLAR consistently outperforms representative baselines. Visualization analyses further show that MOLAR provides interpretable reliability and modality-evidence diagnostics.

Machine-learning emulators are increasingly used for weather prediction and have the potential to extend skill on subseasonal-to-seasonal timescales by learning dynamically important sources of predictability. A key challenge is whether the models can exploit predictability anchors, such as stratospheric variability, that influence tropospheric circulation beyond short lead times. We test how architectural inductive bias affects emulation of sudden stratospheric warming (SSW) dynamics using paired idealised Isca simulations that differ only in an imposed wave-2 heating perturbation. Across convolutional, transformer, and graph-based architectures trained for one-step prediction, model differences are modest when the stratosphere is dynamically quiet but widen substantially when SSW-like variability is active. Our results identify explicit three-dimensional vertical coupling as a key inductive bias for machine-learning emulation of stratospheric dynamics. However, Eliassen-Palm flux diagnostics show that low forecast error does not guarantee physically faithful wave-mean-flow interaction, with coherent errors remaining in stratospheric wave-driving structure.