7 May 2025 in Communications Earth & Environment
Global food security is increasingly challenged by climate change and unsustainable agriculture, emphasizing the need for strategies to enhance crop productivity. Understanding the interplay between crop health and soil microbiomes is crucial. This study explores the link between crop health, observed via multi-spectral satellite imagery, and fungal soil microbiome taxonomy. We associate the normalized difference vegetation index with fungal microbiomes in wheat, barley, and maize using a two-step machine learning process. The first step adjusts normalized difference vegetation index values for abiotic confounders using a random forest model trained on Lucas 2018 topsoil and ERA5 climate datasets. The second step clusters operational taxonomy unit counts from fungal DNA, revealing significant differences in residual normalized difference vegetation index values. To identify potential bio-fertilizer candidates, we compare the average relative abundance of operational taxonomy unit clusters and construct sparse biological networks. Key findings are: (I) clusters with higher plant pathogenic genera have lower normalized difference vegetation index values; (II) clusters with higher influential scores for multiple beneficial genera have higher normalized difference vegetation index values; (III) lower abundance taxonomy (1-3%) seems to regulate microbial networks; (IV) the influence of beneficial vs. pathogenic taxonomy is relative to their abundance. The study links satellite imagery to fungal microbiomes, providing a baseline for exploring fungal bio-fertilizers.
3 March 2025 in Journal of Anxiety Disorders
4 September 2025 in Research Square
Ensuring a stable and nutritious food supply requires sustainable agricultural practices, with soil fungi emerging as promising bio-fertilizers that can enhance crop health. In this study, we assess how the fungal microbiome mediates the relationship between environmental factors and crop health in wheat fields. Using a structural equation model, we analyzed fungal genera together with abiotic variables (weather, soil nutrients, and field type), while crop health was estimated via NDVI values from satellite imagery, providing a cost-effective approach for field monitoring. Our findings from both conventional and ecological wheat fields confirm our hypothesis that the fungal microbiome mediates the effects of environmental factors, such as soil nutrients, seasonality, and weather, on crop health. Notably, we observe a consistent positive association between the fungal genus Mortierella and crop health. Positive associations were also found for Candida, Leucothecium, and Phialemonium, while Thermomyces and Trichocladium correlated negatively. Mortierella had the greatest impact, highlighting its potential as a key biofertilizer.
18 February 2026 in arXiv
Conventional large language model (LLM) fairness alignment largely focuses on mitigating bias along single sensitive attributes, overlooking fairness as an inherently multidimensional and context-specific value. This approach risks creating systems that achieve narrow fairness metrics while exacerbating disparities along untargeted attributes, a phenomenon known as bias spillover. While extensively studied in machine learning, bias spillover remains critically underexplored in LLM alignment. In this work, we investigate how targeted gender alignment affects fairness across nine sensitive attributes in three state-of-the-art LLMs (Mistral 7B, Llama 3.1 8B, Qwen 2.5 7B). Using Direct Preference Optimization and the BBQ benchmark, we evaluate fairness under ambiguous and disambiguous contexts. Our findings reveal noticeable bias spillover: while aggregate results show improvements, context-aware analysis exposes significant degradations in ambiguous contexts, particularly for physical appearance ($p< 0.001$ across all models), sexual orientation, and disability status. We demonstrate that improving fairness along one attribute can inadvertently worsen disparities in others under uncertainty, highlighting the necessity of context-aware, multi-attribute fairness evaluation frameworks.
3 September 2025 in arXiv
HSHarald Vilhelm Skat-Rørdam
Harald Vilhelm Skat-Rørdam
KRKathrine Sofie Rasmussen
Accurate evaluation of event detection in time series is essential for applications such as stress monitoring with wearable devices, where ground truth is typically annotated as single-point events, even though the underlying phenomena are gradual and temporally diffused. Standard metrics like F1 and point-adjusted F1 (F1$_{pa}$) often misrepresent model performance in such real-world, imbalanced datasets. We introduce a window-based F1 metric (F1$_w$) that incorporates temporal tolerance, enabling a more robust assessment of event detection when exact alignment is unrealistic. Empirical analysis in three physiological datasets, two in-the-wild (ADARP, Wrist Angel) and one experimental (ROAD), indicates that F1$_w$ reveals meaningful model performance patterns invisible to conventional metrics, while its window size can be adapted to domain knowledge to avoid overestimation. We show that the choice of evaluation metric strongly influences the interpretation of model performance: using predictions from TimesFM, only our temporally tolerant metrics reveal statistically significant improvements over random and null baselines in the two in-the-wild use cases. This work addresses key gaps in time series evaluation and provides practical guidance for healthcare applications where requirements for temporal precision vary by context.