The SIR model is the cornerstone model for mathematical epidemiology, explaining key epidemic features such as the second-order transition between disease-free and epidemic states, the initial exponential growth of outbreaks or the short-term benefits of control measures. Nonetheless, the classical SIR model assumes that pathogen traits remain fixed, thus neglecting viral evolution. Here we propose a minimal extension of the SIR model, allowing infectiousness to evolve. We show that such evolution can cause superexponential early growth of outbreaks, create abrupt epidemic transitions, and undermine the effectiveness of control policies, as lifting interventions too early can lead to worse epidemic scenarios than no action. We derive analytical expressions for the critical mutation rate and intervention time governing this behavior, and identify a strong asymmetry between control strategies: while shortening the infectious period hinders transmission without suppressing viral evolution, lowering transmission both reduces cases and slows down viral evolution.

This survey paper offers a concise introduction to Kashiwara's work on $\mathcal{D}$-modules, microlocal analysis and related subjects. In this way, we explain his role in the development of algebraic analysis.

The general real singlet extension of the Standard Model (SM), the RxSM, is one of the simplest theories Beyond-the-Standard Model (BSM) that can accommodate a strong first-order electroweak phase transition (SFOEWPT). We investigate the possible thermal histories of the scalar potential in the RxSM, and the regions of the model parameter space in which SFOEWPT can be realised. We then explore complementary avenues to probe such scenarios experimentally: either using searches for a stochastic background of gravitational waves (GWs), or using searches for di-Higgs production processes at future collider experiments, focusing on the case of a high-energy $e^+e^-$ collider. An important aspect of our work is that one-loop corrections to all relevant trilinear scalar couplings are consistently included both in the calculation of dynamics of the electroweak phase transition (EWPT) and in collider processes. We find entirely different phenomenological signatures for different parts of the RxSM parameter space giving rise to SFOEWPTs. On the one hand, if the SFOEWPT is driven by the singlet field, the 125 GeV Higgs boson is very SM-like and signs of BSM physics would be difficult to find at colliders, but strong GW signals could be produced. On the other hand, in scenarios where a SFOEWPT is driven by the doublet field, BSM deviations in properties of the detected Higgs boson, particularly in its trilinear self-coupling, typically lead to observable signals at colliders, while detectable GW signals are much more challenging to achieve. This work highlights the complementarity of collider experiments and cosmological observations to determine the dynamics of the EWPT and reconstruct the shape of the Higgs potential realised in Nature.

Understanding anharmonicity is crucial for designing materials with desired lattice thermal conductivity. Designing a material descriptor that effectively captures anharmonicity while being cost-effective remains a significant challenge. This work proposes a simple metric that helps explain the diversity in lattice thermal conductivity (kl) among materials by quantifying their anharmonic effects. This descriptor "phonon band center" (PBC) encapsulates the critical factors associated with the physics of phonon scattering, revealing a simple inverse relationship with the Gruneisen parameter, the response of phonons with changing volume, and strong correlation with lattice thermal conductivity. This metric has been established using the chalcopyrite class of materials and subsequently validated across various classes of materials using experimental kl. Our approach effectively differentiates materials based on PBC, thereby streamlining the identification of candidates with desirable kl.

For finite-dimensional algebras over a field, Koenig and Yang established a bijection between silting complexes and simple-minded collections in the bounded derived category, with further contributions by many authors in various settings. In this paper, we work over a commutative complete local noetherian ring $(R,\m,k)$ rather than over a field and establish a bijection in this more general setting. As an application of this generalization, we construct a bijection between silting complexes over a noetherian $R$-algebra $Λ$ and silting complexes over $Λ\ten^\LL_RS$ for any morphism of commutative complete local noetherian rings $(R,\m,k)\to(S,\n,k)$. This result generalizes some known results on silting complexes over noetherian algebras.

The security of modern JavaScript (JS) engines is critical since they provide the primary defense mechanism for executing untrusted code on the web. The recent integration of WebAssembly (Wasm) has transformed these engines into complex polyglot environments, creating a novel attack surface at the JS-Wasm interaction boundary due to the distinct type systems and memory models of two languages. This boundary remains largely underexplored, as previous works mainly focus on testing JS and Wasm as two isolated entities rather than investigating the security implications of their cross-language interactions. This paper proposes Weaver, an effective greybox fuzzing framework specifically tailored to uncover vulnerabilities at the JS-Wasm boundary. To comply with the language constraints, Weaver uses a type-aware generation strategy, meticulously maintaining the dual-type representation for every generated variables. This allows fuzzer to validly utilize variables across the language boundary. Besides, Weaver leverages the UCB-1 algorithm to intelligently schedule mutators and generators to maximize the discovery of new code paths. We have implemented and evaluated Weaver on three JS engines. The results indicate that Weaver achieves superior code coverage compared to state-of-the-art fuzzers. Moreover, Weaver has uncovered two new bugs in the latest versions of these engines, one of which is considered high severity and set to highest priority, demonstrating the practicality of Weaver.

The standard approach to quantum measurements is to assume that they lead to effectively instantaneous collapse of the quantum state. However, if we assume that we are unable to enforce at what exact moment of time the measurement occurs due to a finite resolution of any time measurement device, at the level of the ensemble, the measurement would lead to an effectively nonunitary evolution involving a mixed state. Each individual ensemble member would face an instantaneous collapse at different moments of time. This process is completely indistinguishable from fundamental nonunitary evolution at the level of each individual ensemble member, within the framework of strong projective measurements. In this paper, we show that weak postselected measurements can distinguish these two types of evolution. An experimental protocol for determining the nature of quantum collapse is described, and the example of a hydrogen atom is analyzed in detail.

In this paper, we show a stochastic approach to generate a 3D model of a foliage, which is then used for deterministic ray-tracing channel modeling. This approach is verified by a measurement campaign at 60 and 80 GHz with 2 GHz bandwidth. The wireless channel is characterized by path-loss and RMS delay spread and we show the angular dependency of those parameters when the receiver is placed on a half-circle around the tree. Besides electromagnetic material properties, the 3D model is characterized by several interpretable parameters, including tree volume, leaf size, leaf density, and the tree crown shape parameter.

We prove that the index of a CMC surface with capillary boundary is bounded from above linearly by its genus, number of boundary components, and branching order, and also by some Willmore-type energy involving the area, mean curvature, contact angle, and ambient curvature. The main auxiliary theorem of more general interest is a comparison of the second variations of area and energy at a branched conformal map with boundary. In the appendix we derive the various second variation formulae for area, enclosed-volume, and wetting functionals away from critical points and for non-admissible variations, the purpose of which is to rather comprehensively fill a gap in the literature.

We report a comparative study on transmon qubit control using (i) conventional attenuated coaxial microwave line and (ii) an optical control system using modulated laser light delivered over telecommunications optical fiber to a photodiode located at the 1K stage of a dilution cryostat. During each experiment, we performed repeated measurements of the energy relaxation and coherence times of a transmon qubit using one of the control signal delivery methods. Each measurement run spanned 20 hours of measurement time and from these datasets we observe no measurable effect on coherence of the qubit compared to random coherence fluctuations. Our results open up the possibility of large scale integration of the optical qubit control system.

The integration of diverse quantum resources and the exploitation of more degrees of freedom provide key operational flexibility for universal fault-tolerant quantum computation. In this work, we propose a flexible Gottesman-Kitaev-Preskill-state-embedded fault-tolerant quantum computation architecture based on a three-dimensional cluster state constructed in polarization, frequency, and orbital angular momentum domains. Specifically, we design optical entanglement generators to produce three diverse entangled pairs, and subsequently construct a three-dimensional cluster state via a beam-splitter network with several time delays. Furthermore, we present a partially squeezed surface-GKP code to achieve fault-tolerant quantum computation and ultimately find the optimal choice of implementing the squeezing gate to give the best fault-tolerant performance (the fault-tolerant squeezing threshold is 11.5 dB). Our scheme is flexible, scalable, and experimentally feasible, providing versatile options for future optical fault-tolerant quantum computation architecture.

The Improved Partial Area-Analytical Calculation (IPA-AC) method represents a leading meshfree discretization strategy for peridynamic models, distinguished by its rigorous geometric treatment of boundary intersections via dual corrections of integration weights and quadrature points. Despite its empirical success in suppressing boundary-induced geometric errors, a systematic theoretical characterization of its convergence behaviors under distinct scaling limits has remained elusive. This work establishes a unified convergence framework for the IPA-AC method applied to both scalar and tensor kernels. By leveraging the Lax Equivalence Theorem, we explicitly derive error estimates that reveal the method's performance across three critical limiting regimes. The theoretical analysis, substantiated by numerical validation, demonstrates that: (1) for a fixed horizon $δ$, the method achieves robust second-order convergence $\mathcal{O}(h ^{2})$ with respect to the mesh size $h$; (2) for a fixed mesh, the discretization error scales as $\mathcal{O}(δ^{-2})$, indicating a sensitivity to the nonlocal length scale; and (3) the method does not satisfy the Asymptotic Compatibility (AC) condition. These findings clarify that while the IPA-AC method offers superior accuracy for simulating fixed nonlocal models, it requires a sufficiently large horizon-to-mesh ratio to mitigate intrinsic discretization errors when approximating the local limit.

We propose a physics-grounded mechanism design for dynamic spectrum sharing that bridges the gap between radiometric retrieval constraints and economic incentives. We formulate the active and passive users coexistence problem as a Vickrey-Clarke-Groves (VCG) auctions mechanism, where the radiometer dynamically procures ``quiet'' time-frequency tiles from active users based on the marginal reduction in retrieval error variance. This approach ensures allocative efficiency and dominant-strategy incentive compatibility (DSIC). To overcome the computational intractability of exact VCG on large grids, we derive an approximation algorithm by using the monotone submodularity induced by the radiometer equation. AMSR-2-based simulations show that the approach avoids high-cost tiles by aggregating low-cost spectrum across time and frequency. In an interference-trap case study, the proposed framework reduces procurement costs by about 60% over a fixed-band baseline while satisfying accuracy targets.

$\require{mediawiki-texvc}$Galaxy assembly was already well underway in the first 400 Myr of cosmic time, as recently revealed by JWST. However, the contribution of these early galaxies to cosmic reionisation remains uncertain. Here we present new JWST/NIRSpec observations of GS-z13-1-LA obtained as part of the OASIS and JADES programmes, whose combined deep (56 h) NIRSpec/PRISM spectrum confirms the Lyman-$\mathrmα$ line detection and blue UV continuum at redshift $z = 13.1$ presented in a previous work. The measured Lyman-$\mathrmα$ emission (rest-frame equivalent width of $66_{-9}^{+10}\,Å$) and steep continuum slope ($β_\text{UV} \approx -3$) point towards GS-z13-1-LA hosting a remarkably hot and powerful ionising source, and allow at most a modest contribution from the nebular continuum. The steep turnover of the continuum is still present, but less pronounced in the new OASIS spectrum. Combined, this implies that ionising photons may escape GS-z13-1-LA at a sufficient rate to weaken the other, still undetected UV lines, and to lead the formation of a small ionised bubble ($R_\text{ion} \approx 0.2\,\mathrm{pMpc}$). A yet larger bubble could alleviate the required ionising production efficiency of GS-z13-1-LA from $ξ_\mathrm{ion} \approx 10^{26.4}\,\mathrm{Hz\,erg^{-1}}$ down to $\approx 10^{25.9}\,\mathrm{Hz\,erg^{-1}}$, still extremely high but more readily reconcilable with stellar models. In turn, this would require a notable overdensity of galaxies with highly efficient ionising capabilities, a scenario for which tentative evidence is found in the form of 16 nearby photometric candidates and one spectroscopically confirmed source, JADES-GS-z13-0. The new OASIS observations therefore confirm the overall picture of GS-z13-1-LA as an early beacon of reionisation, providing compelling evidence for its start only 330 Myr after the Big Bang.

We consider a cross-diffusion system for which the diffusion of each species is governed solely by the aggregate density through a pressure law of logarithmic or fast diffusion type. The model is set over a one dimensional bounded interval, equipped with no-flux boundary conditions, and accommodates for the presence of potential drifts which are allowed to differ across each species. We establish the global existence of solutions without having to assume the total mixing of solutions. As a consequence, we give a full resolution of the PDE systems recently studied by the authors and by Elbar--Santambrogio, by allowing a general class of initial data with finite bounded variation, with no further structural assumptions on their supports.

We present a systematic classification of uncolored bonded knots with singularity number at most seven. Bonded knots provide a topological model for closed protein chains with intramolecular bridges, such as disulfide bonds. Following the tradition of knot tabulation, we describe a procedure based on the generation of planar graphs, their conversion into bonded knot diagrams, and the use of the Yamada polynomial together with brute-force Reidemeister moves to distinguish topological knotted types.

In this paper, we establish the triality twisted trace formula for PGSO(8), including its discrete part, and obtain a coarse classification of its automorphic representations by combining the properties of triality. By comparing the standard trace formula for G_2 with the triality twisted trace formula for PGSO(8), we derive a corresponding coarse classification for automorphic representations of G_2. Specifically, we construct the triality-twisted elliptic endoscopic data for PGSO(8), and the elliptic endoscopic data for G_2. Based on these constructions and the general framework of trace formulas, we establish the relevant trace formulas. Utilizing the triality property of PGSO(8), we obtain a coarse classification of its automorphic representations, which in turn yields a coarse classification for those of G_2.

Atmospheric turbulence severely limits the coupling of received optical wavefronts into single-mode fibers in satellite-to-ground free-space optical links. Spatial demultiplexing receivers address this challenge by distributing the incoming field across a bundle of single-mode fibers whose outputs are recombined coherently, relaxing the requirements on wavefront correction. In this work, we investigate the design of such receivers from two complementary angles. We first compare the power coupling statistics achieved by several modal bases and show that the spatial support of the modes matters far more than the specific choice of basis, questioning the relevance of mode-selective approaches for this application. We then present the inverse design of a compact two-plane refractive system optimized directly over an ensemble of turbulence realizations using stochastic gradient descent, with no constraint imposed on the input modal decomposition. The optimized system significantly improves over direct coupling into the fiber bundle, approaches the performance of an ideal modal projection, and remains competitive across a broad range of turbulence conditions.

Mathematical modeling offers a valuable approach to understanding Alzheimers disease (AD) given its complexity, unknown causes, and lack of effective treatments. Models, once validated, offer a powerful tool to test medical hypotheses that are otherwise difficult to verify directly. Our focus here is on elucidating the spread of misfolded tau protein, a critical hallmark of AD alongside Abeta protein, taking also into account the synergistic interaction between the two proteins. We consider distinct modelling choices, all employing network frameworks for protein evolution, differentiated by their network architecture and diffusion operators. By carefully comparing these models against clinical tau concentration data, gathered through advanced multimodal analysis techniques, we show that certain models replicate better the proteins dynamics. This investigation underscores a crucial insight: in modeling complex pathologies, the precision with which the mathematical framework is chosen is crucial, especially when validation against clinical data is considered decisive.

Germanium-silicon-germanium (Ge/Si$_{x}$Ge$_{1-x}$) heterostructures have emerged as a promising platform for hole-spin quantum technologies and high-mobility electronics, where strain and quantum confinement strongly reshape the Ge valence bands. However, the momentum-resolved valence-band structure of buried strained Ge quantum wells has so far been inferred only indirectly. Here we use soft X-ray angle-resolved photoemission spectroscopy (SX-ARPES) to directly probe the electronic structure of strained Ge quantum wells embedded in SiGe barriers. We resolve strain-split and size-quantized valence subbands, determine their heavy-hole, light-hole and split-off composition, and measure the valence-band offset at the Ge/SiGe heterojunction. Comparison with ab initio calculations shows that an accurate description requires explicit inclusion of the confinement potential imposed by the SiGe barrier, which plays a decisive role in determining the dispersion, ordering and mixing of the hole states. Our results provide the first direct experimental picture of how strain and confinement determine the valence-band structure of Ge quantum wells, establishing a foundation for predictive modelling of hole-spin qubits and high-mobility devices based on group-IV heterostructures.

In this article, we investigate when the ordinary and symbolic powers of the Alexander dual of connected ideals of graphs coincide, and provide a complete classification of all such graphs. Furthermore, we prove Conforti--Cornuèjols conjecture for this class of ideals.

The study of spontaneous supersymmetry breaking (SSB) on the lattice is obstructed by a severe sign problem. Quantum computing provides a promising alternative approach. In particular, properties of supersymmetry relate SSB to the ground-state energy, which can be probed using hybrid quantum--classical algorithms such as the variational quantum eigensolver (VQE). In this work we present VQE analyses for supersymmetric quantum mechanics with various superpotentials. A key new feature is an adaptive ansatz construction algorithm that reduces the number of variational parameters within our ansätze. This lowers the resource burden on both the classical optimizer and the noisy quantum processor, thereby improving the feasibility of these calculations in the NISQ era. Additionally, we present preliminary VQE results obtained from real IBM quantum devices, highlighting accuracy, resource constraints, and computational cost, both with and without the application of error mitigation techniques.

Local composition fluctuations in random alloys become crucial when one or more dimensions are reduced to the nanoscale. Using extended Hückel theory, we study the semiconductor random alloy SiGe sandwiched between Si due to its relevance for transistor devices. We evaluate the effects of the alloy composition, layer thickness, and local fluctuations of the Ge concentration on the band alignment and the band gap. The results are compared with the finite quantum well model. That model captures the essential physics and can act as a computationally faster alternative.

We consider semilinear parabolic optimal control problems subject to Neumann boundary conditions, control constraints, and an infinite time horizon. The control constraints are pointwise in time, but they can be pointwise or integral in the space variable. Crucially, the optimal control problem does not include a Tikhonov regularization in the cost functional, which provides a major difficulty in the extension of the classical finite-horizon theory to infinite-horizon optimal control problems. As a consequence of our findings, we establish a sufficient second-order optimality condition and prove that local optimal states of the finite-horizon problems approximate local optimal states to the infinite-horizon problem as the horizon tends to infinity.

Context. The High Resolution Telescope of the Polarimetric and Helioseismic Imager on Solar Orbiter (SO/PHI-HRT) operates in an extreme observational environment, observing the Sun as close as $0.28$ au. The high thermal load and large illuminating field puts high demands on the instrument in terms of both imaging performance and false light control. Aims. To characterise the amount of stray light (false light) within SO/PHI-HRT, apply a correction, and re-compare the data products with the Helioseismic and Magnetic Imager on the Solar Dynamics Observatory (SDO/HMI). Methods. We analyse solar limb profiles and a Mercury transit to quantify the amount of stray light and add a correction term when partially reconstructing the SO/PHI-HRT images. For the comparison with SDO/HMI we use data from the 2023 March Solar Orbiter inferior conjunction and compare the magnetic fields on a pixel-by-pixel basis. Results. Increased continuum intensity contrast in the quiet Sun, and darker intensity levels are found in strong magnetic features. Consequently, much stronger fields are inferred in these features. Comparing the stray light corrected data with that from the standard SDO/HMI data products results in a much closer agreement across all vector magnetic field components, particularly when the cadence and noise levels are identical. In most solar features, SO/PHI-HRT infers stronger fields than the SDO/HMI line-of-sight magnetograms. Compared to the vector magnetic field from SDO/HMI the two are very well aligned, with only slight differences in the strongest field regions (where $|\mathbf{B}|>1600$ G or $|\mathbf{B_{\text{LOS}}}|>1300$ G).

As the landscape of software engineering evolves, introductory programming courses must go beyond teaching syntax to foster comprehensive technical competencies and professional soft skills. This paper reports on a pedagogical experience in a "Fundamentals of Programming" course that used a Project-Based Learning (PBL) framework to develop a 2D "Maze Runner"-style game. While game development serves as a high-engagement vehicle for mastering core concepts, such as multidimensional arrays, control structures, and logic, the core of this study focuses on implementing a rigorous, multifaceted assessment model structured across four distinct dimensions: (1) an in-situ technical demonstration, evaluating real-time code execution and algorithmic robustness; (2) a technical screencast, requiring students to articulate their work in a concise audiovisual format; (3) a formal presentation to instructors, defending their project's design patterns and problem-solving strategies; and (4) a structured peer-review process, where students evaluated their colleagues' projects. Our findings suggest that this multi-dimensional approach not only improves student retention of programming fundamentals but also significantly enhances communication skills and critical thinking. By integrating peer evaluation and multimedia documentation, the course successfully bridges the gap between basic coding and the collaborative requirements of modern software engineering. This paper details the curriculum design, the challenges of implementing diverse assessment pillars, and the measurable impact on student performance and engagement, providing a scalable roadmap for educators looking to modernize introductory computing curricula.

There is now strong evidence that the Milky Way (MW) hosts a nuclear stellar disc (NSD). However, whether the NSD is purely axisymmetric or contains a nuclear bar remains unresolved. Since approximately $50\%$ of barred galaxies with MW-like mass in the local Universe host a nuclear bar, investigating whether the MW hosts one is of interest. We conduct a systematic analysis to identify robust kinematic diagnostics capable of determining whether the MW hosts a nuclear bar. Using N-body simulations, we explore the kinematic signatures indicative of a nuclear bar. Using the phase-space coordinates longitude $(\ell)$, latitude $(b)$, proper motions ($μ_\ell$ and $μ_{\rm b})$ and line-of-sight velocity $(v_{\rm los})$, we test various diagnostics assuming different nuclear bar orientations. We also evaluate how sample size, dust extinction and bar amplitude influence the efficacy of the diagnostics. We identify two independent kinematic diagnostics capable of revealing a nuclear bar in the MW: (1) the vertex deviation, $l_{\rm v}$, of the ($v_{\ell}-v_{\rm los}$) velocity ellipse; and (2) The asymmetry in the $μ_{\ell}$ vs $\ell$ distribution. While both are impacted by the sample size and extinction, the vertex deviation proves more robust, especially when combining stars from multiple observational fields. We also assess the correlation between the line-of-sight velocity and the $h_3$ Gauss-Hermite moment ("skewness") of the line-of-sight velocity but find no clear distinction between an NSD and a nuclear bar based on this metric. Our results suggest that data from the current KMOS survey may allow a marginal detection of a nuclear bar using the vertex deviation method. A companion paper provides further validation and detailed analysis of this approach. Nonetheless, future surveys will provide the high quality data necessary to fully exploit the diagnostics outlined in this study.

We leveraged reliable age and distance estimates from LAMOST-DR3 and APOGEE-DR17+AstroNN combined with \gaia\ data to perform a detailed analysis of the stellar age distribution in the Milky Way's (MW) outer disc using giant stars. Selecting stars near the midplane ($|z|<0.3$ kpc) on near-circular orbits ($λ_c > 0.9$), we analysed these independent datasets that employed different age-estimation methods. Our stringent kinematic selection criteria effectively exclude halo stars, ensuring that the observed age trends reflect genuine disc properties rather than contamination from older halo populations. Our results reveal a 'U-shaped' stellar age profile, where a negative gradient in the inner disc transitions to a positive gradient in the outer disc region. We identify the minimum in the stellar age profile at $R_{\rm min}=11.28 \pm 0.58$ kpc and $R_{\rm min}=12.15\pm 0.62$ kpc for the APOGEE-DR17 and LAMOST-DR3 samples, respectively. Using N-body+SPH simulations, we demonstrate that $R_{\rm min}$ corresponds to the break radius in the stellar density profile ($R_{\rm br}$), marking the edge of the Galaxy's star-forming disc. This break arises from a sharp decline in the star formation rate, with the outer positive age gradient produced by the radial migration of stars born inside $R_{\rm br}$. The cessation of star formation in the outer disc might be due to several mechanisms, including the dynamical influence of the bar's outer Lindblad resonance, the onset of the Galactic warp, or thermally regulated star formation. Overall, our results support the picture that the MW has a Type II (down-bending) stellar disc with a break at $R_{\rm br} \approx 11.28-12.15$ kpc, where the combination of star-formation cut-off and radial migration produces the observed U-shaped age profile.

Programming environments typically separate the world of static code from the dynamic execution of programs. Developers must switch between writing code and observing its execution, often with limited tools to understand the relationship between code changes and runtime behavior. Several paradigms and approaches exist to bridge this gap, including exploratory programming for comparing code variants, live programming for immediate feedback, and omniscient debugging for exploring execution history. However, existing solutions tend to focus on specific aspects and one specific paradigm rather than providing a fully integrated environment with multiple capabilities. This paper introduces \spacetime Programming, a novel approach that unifies these paradigms to create a programming model for exploring both code modifications and execution flow. At the core of our approach is a trace mechanism that captures not only execution state but also the corresponding code changes, enabling developers to explore programs in both space (code variants) and time (execution flow). As a proof of concept, we implemented a Python library supporting SpaceTime Programming and applied it in two contexts: a live omniscient debugger and a Pygame game development tool, showcased through a Flappy Bird-like game. We further evaluated SpaceTimePy on five real-world Python projects, finding performance overhead ranging from 35% to 150% on test suites.

Context: The increasing adoption of machine learning (ML) and artificial intelligence (AI) technologies raises growing concerns about their environmental sustainability. Developing and deploying ML-enabled systems is computationally intensive, particularly during training and inference. Green AI has emerged to address these issues by promoting efficiency without sacrificing accuracy. While prior research has proposed catalogs of sustainable practices (i.e., green tactics), there remains limited understanding of their adoption in practice and whether additional, undocumented tactics exist. Objective: This study aims to investigate the extent to which existing sustainable practices are implemented in real-world ML-enabled systems and to identify previously undocumented practices that support environmental sustainability. Method: We conduct a mining software repository study on 205 open-source ML projects on GitHub. To support our analysis, we design a novel mechanism based on large language models (LLMs) capable of identifying both known and new sustainable practices from code repositories. Results: Our findings confirm that green tactics reported in the literature are used in practice, although adoption rates vary. Furthermore, our LLM-based approach reveals nine previously undocumented sustainable practices. Each tactic is supported with code examples to aid adoption and integration. Conclusions: We finally provide insights for practitioners seeking to reduce the environmental impact of ML-enabled systems and offer a foundation for future research in automating the detection and adoption of sustainable practices.