Large language models (LLMs) for code completion and generation are increasingly used in software development, yet they may reproduce training examples verbatim and without authorship attribution, raising legal and ethical concerns around plagiarism and license compliance. Classical fingerprint-based plagiarism detectors, such as Winnowing, remain highly effective, yet the inspection requires comparing fragments of code to the entire training set, and their linear-time search makes them impractical for the billion-scale corpora used to train modern code LLMs. To bridge this gap, we introduce SourceTracker, a 300M-parameter encoder tailored for code retrieval, together with a hybrid two-stage provenance-tracking pipeline HybridSourceTracker (HST). HST first narrows down a small set of candidate snippets via vector search, then re-ranks those candidates using Winnowing on exact fingerprints. We train and evaluate our system on a 10M-snippet subset of the TheStackV2 dataset, with both verbatim and adapted snippets that emulate realistic identifier renaming. On an in vitro 100k-snippet search space with adapted queries, our hybrid approach reaches a mean reciprocal rank on par with Winnowing for 30-token fragments. Then, starting from windows ≥ 60 tokens, it consistently over-performs by up to 5.4% while preserving logarithmic-time query complexity. In a complementary evaluation using an LLM-based judge, we find that many retrieved snippets not labeled as ground truth are still highly similar to the expected sources, particularly with longer context windows, and thus remain useful for end users. Overall, our results demonstrate that integrating vector search with fingerprinting enables scalable, high-precision provenance tracking for code produced by LLMs, provided that their training data is made accessible. More broadly, our findings underscore the growing need for transparency around LLM training corpora as a prerequisite for effective provenance tracking, attribution, and license compliance.
Sorption heat storage allows solar thermal energy to be stored for weeks or months with negligible losses, making it a promising route to decarbonize space heating in buildings, yet most sorbent data are acquired under closed-system conditions that poorly represent the open, atmospheric-pressure operation of solar-driven building-scale stores. Here, we report the thermal characterization, under open-system boundary conditions (Tdes = 80 °C, Tads = 20 °C, pads = 11.6 mbar), of the amino-functionalized titanium framework NH2-MIL-125(Ti) and of two composites, namely, NH2-MIL-125(Ti)+45 wt% CaCl2 and NH2-MIL-125(Ti)+45 wt% SrBr2. To the best of our knowledge, NH2-MIL-125(Ti)/SrBr2 is a new MOF/salt pairing, and no NH2-MIL-125(Ti)/salt composite has previously been characterized under open-system conditions. Materials were characterized by X-ray powder diffraction, N2 physisorption (BET) and simultaneous thermogravimetry–differential scanning calorimetry (TG–DSC) coupled to a controlled-humidity generator. The pristine framework delivered a water uptake of 0.21 g g−1 and an energy storage capacity of 163 Wh kg−1, whereas the salt loading increased these values to 0.56 g g−1 and 342 Wh kg−1 (CaCl2) and 0.40 g g−1 and 313 Wh kg−1 (SrBr2), respectively, at the expense of slower kinetics. All the materials retained their water uptake over six adsorption/desorption cycles (uptake loss ≤5%; energy storage capacity loss 8–12%, reflecting a slight decrease in the measured heat of adsorption). A quantitative kinetic analysis (characteristic times t50/t90 and linear-driving-force constants) is provided. The results establish NH2-MIL-125(Ti)/salt composites as promising sorbents for low-temperature solar heat storage, the 80 °C regeneration temperature being directly compatible with common flat-plate solar collectors, and provide a rare open-system dataset for this material class.
Abstract In September 2021, the European Union launched the New European Bauhaus (NEB) initiative to promote a green, inclusive, and aesthetically oriented transition across Europe. Although the NEB emphasizes sustainability, inclusiveness, and aesthetics, existing assessment frameworks rarely integrate these dimensions in a cohesive or adaptable manner. The objective of this research is to establish a flexible component-based framework capable of quantitatively capturing the core concepts of the NEB. The framework is developed through an extensive literature review and organized into a hierarchical structure that aggregates fine-grained performance metrics into dimension indices and a final composite performance index. A fuzzy Technique for Order Preference by Similarity to Ideal Solution (TOPSIS)–based weighting mechanism is incorporated to enhance adaptability and to better reflect the inclusive and participatory ethos of the NEB. The framework’s applicability is demonstrated through its use on the Mjøstårnet timber tower in Norway. Following an initial performance evaluation, the tower is redesigned using adaptive control principles and reassessed to demonstrate the framework’s responsiveness to design modifications and its capacity to capture multidimensional performance. The results indicate that the proposed framework can function as a robust and versatile tool to support policymaking, design innovation, and performance evaluation aligned with the evolving priorities of European construction.
Glioblastoma Multiforme (GBM) is a highly aggressive brain tumour with limited therapeutic options and poor prognosis. This study presents a spatially resolved, nonlinear mathematical framework to investigate the efficacy of immunotherapy strategies based on cytotoxic T-lymphocyte (CTL) infusion. The model couples tumour and immune dynamics through a system of partial differential equations (PDEs), incorporating cell proliferation, diffusion, and chemotactic migration in response to TGF-β, a tumour-secreted signalling molecule. A reduced ordinary differential equation (ODE) model is first analysed to derive threshold conditions for tumour eradication, identifying critical infusion levels consistent with clinical data. Numerical bifurcation analysis further reveals the influence of key model parameters on tumour persistence and therapeutic response. The full PDE model is solved using the finite element method on simplified 2D domains, followed by sensitivity analyses to quantify parameter influence on tumour mass and volume. The model is then applied to a realistic 3D brain geometry reconstructed from patient-specific MRI and DTI data, accounting for anatomical anisotropy and tissue heterogeneity. Therapeutic scenarios are simulated with spatially localised lymphocyte infusion. Results show the complex spatial interplay between tumour growth, immune response, and brain microstructure, showing that infusion intensity and tumour location critically influence therapeutic outcomes. These findings highlight the importance of incorporating patient-specific anatomical features and spatial heterogeneity into spatially informed mathematical models to support the design and optimisation of personalised immunotherapy strategies for GBM.
Spray-based 3D concrete printing (S-3DCP) offers advantages in formwork-free fabrication, multidirectional deposition, and reinforcement integration, yet material–process–structure synergy and engineering evaluation remain insufficiently understood. Taking this synergy as the central framework, this review synthesizes S-3DCP studies published up to 2026 and establishes a technological framework covering material design, process control, reinforcement integration, and engineering applications. The results show that stable fabrication depends on the coordinated matching of material properties with equipment and process conditions, while interlayer interface formation and reinforcement encapsulation affect structural performance. Future research should develop transferable evaluation frameworks and closed-loop control methods and strengthen full-scale structural and long-term service validation.
