Assembly Theory provides a promising framework to explain the complexity of systems such as molecular structures and the origins of life, with broad applicability across various disciplines. In this study, we explore and consolidate different aspects of Assembly Theory by introducing a simplified Toy Model to simulate the autocatalytic formation of complex structures. This model abstracts the molecular formation process, focusing on the probabilistic control of catalysis rather than the intricate interactions found in organic chemistry. We establish a connection between probabilistic catalysis events and key principles of Assembly Theory, particularly the probability of a possible construction path in the formation of a complex object, and examine how the assembly of complex objects is impacted by the presence of autocatalysis. Our findings suggest that this presence of autocatalysis tends to favor longer consecutive construction sequences in environments with a low probability of catalysis, while this bias diminishes in environments with higher catalysis probabilities, highlighting the significant influence of environmental factors on the assembly of complex structures.

Ensuring reliability, availability, and security in modern software systems hinges on early fault detection, yet predicting which parts of a codebase are most at risk remains a significant challenge. In this paper, we analyze 2.4 million commits drawn from 33 heterogeneous open-source projects, spanning healthcare, security tools, data processing, and more. By examining each repository per file and per commit, we derive process metrics (e.g., churn, file age, revision frequency) alongside size metrics and entropy-based indicators of how scattered changes are over time. We train and tune a gradient boosting model to classify bug-prone commits under realistic class-imbalance conditions, achieving robust predictive performance across diverse repositories. Moreover, a comprehensive feature-importance analysis shows that files with long lifespans (high age), frequent edits (revision count), and widely scattered changes (entropy metrics) are especially vulnerable to defects. These insights can help practitioners and researchers prioritize testing and tailor maintenance strategies, ultimately strengthening software dependability.

Phishing websites pose a major cybersecurity threat, exploiting unsuspecting users and causing significant financial and organisational harm. Traditional machine learning approaches for phishing detection often require extensive feature engineering, continuous retraining, and costly infrastructure maintenance. At the same time, proprietary large language models (LLMs) have demonstrated strong performance in phishing-related classification tasks, but their operational costs and reliance on external providers limit their practical adoption in many business environments. This paper presents a detection pipeline for malicious websites and investigates the feasibility of Small Language Models (SLMs) using raw HTML code and URLs. A key advantage of these models is that they can be deployed on local infrastructure, providing organisations with greater control over data and operations. We systematically evaluate 15 commonly used SLMs, ranging from 1 billion to 70 billion parameters, benchmarking their classification accuracy, computational requirements, and cost-efficiency. Our results highlight the trade-offs between detection performance and resource consumption. While SLMs underperform compared to state-of-the-art proprietary LLMs, the gap is moderate: the best SLM achieves an F1-score of 0.893 (Llama3.3:70B), compared to 0.929 for GPT-5.2, indicating that open-source models can provide a viable and scalable alternative to external LLM services.

Grassland biomass yield reflects a complex interaction of management intensity and environmental factors, yet quantifying the relative role of practices such as mowing and fertilization remains challenging. In this study, we introduce a multi-class machine learning framework to predict above-ground biomass on 150 permanent grassland plots across eight years (2009–2016) in Germany’s Biodiversity Exploratories and to evaluate the influence of key management variables. Following rigorous data cleaning, imputation of missing nitrogen values, feature standardization, and encoding of categorical practices, we trained CatBoost classifiers optimized via Bayesian hyperparameter search and mitigated class imbalance with ADASYN oversampling. We assessed model performance under binary, three-class, four-class, and five-class quantile-based categorizations, achieving test accuracies of 0.76, 0.57, 0.42, and 0.38, respectively. Across all schemes, mowing frequency and mineral nitrogen input emerged as the dominant predictors, while secondary variables such as drainage and conditioner use contributed as well. These results demonstrate that broad biomass categories can be forecast reliably from standardized management records, whereas finer distinctions necessitate additional environmental information or automated sensing to capture nonlinear effects and reduce reporting bias. This work shows both the potential and the limits of machine learning for informing sustainable grassland management and explainability thereof. Frequent mowing and higher mineral nitrogen inputs explained most of the predictable variation, enabling a 76% accurate separation of low and high biomass categories. Predictive accuracy fell below 60% for finer class resolutions, indicating that management records alone are insufficient for detailed yield forecasts without complementary environmental data.

Grassland biomass yield reflects a complex interaction of management intensity and environmental factors, yet quantifying the relative role of practices such as mowing and fertilization remains challenging. In this study, we introduce a multi‑class machine learning framework to predict aboveground biomass on 150 permanent grassland plots across eight years (2009–2016) in Germany’s Biodiversity Exploratories and to evaluate the influence of key management variables. Following rigorous data cleaning, imputation of missing nitrogen values, feature standardization, and encoding of categorical practices, we trained CatBoost classifiers optimized via Bayesian hyperparameter search and mitigated class imbalance with ADASYN oversampling. We assessed model performance under binary, three‑class, four‑class, and five‑class quantile‑based categorizations, achieving test accuracies of 0.76, 0.57, 0.42, and 0.38, respectively. Across all schemes, mowing frequency and mineral nitrogen input emerged as the dominant predictors, while secondary variables such as drainage and conditioner use contributed as well. These results demonstrate that broad biomass categories can be forecast reliably from standardized management records, whereas finer distinctions necessitate additional environmental information or automated sensing to capture nonlinear effects and reduce reporting bias. This work shows both the potential and the limits of machine learning for informing sustainable grassland management and explainability thereof.