This paper uses machine learning to figure out the most effective ways to upgrade thousands of homes in a city, like insulating roofs or walls, to save energy and cut carbon emissions, while also estimating how certain those savings are.
Fabric retrofit scenario analysis under uncertainty for an urban housing stock using machine learning · Yunfei Bai, Aaron Tesfa Tsion, Raúl Rosales et al. · arXiv:2610.24608
This paper forecasts future carbon emissions from commercial buildings, like offices and stores, across 16 key countries until 2060, identifying where emissions are likely to peak and the biggest opportunities to cut them.
Commercial building decarbonization across countries: Future emission trajectories and mitigation priorities · Qinwen Tang, Tianyi Wang, Nan Zhou et al. · arXiv:2610.26696
A model to understand how transportation networks can experience widespread collapse, not just local disruptions, when faced with climate-related events like heavy rainfall. Even localized rain can trigger system-wide failures if it crosses a critical intensity, revealing hidden vulnerabilities.
Systemic collapse emergence from climate-induced perturbations in transportation networks · Mauro Faccin, Tomas Scagliarini, Manlio De Domenico · arXiv:2610.26428
A rigorous way to understand when complex systems, like power grids, can maintain stable synchronization. Precise boundaries for how much we can trust simplified mathematical models, ensuring that engineers know exactly when these approximations accurately predict stable operation.
Linear Approximations Provide Rigorous Conditions for Phase Locking in Oscillator Networks and Power Grids · Iva Bačić, Carsten Hartmann, Philipp C. Böttcher et al. · arXiv:2610.24595
A look at how to best translate market predictions into actual trades, focusing on designing optimal strategies that account for real-world trading costs and the temporary price impact of your actions.
Optimal Trading Filters and the Value of Predictive Signals · Matteo Smerlak · arXiv:2610.24558
This paper helps us find meaningful groups, or "communities," within complex networks like social media or biological systems, but with a crucial difference: it lets us build in real-world rules and prior knowledge.
Constrained modularity maximization in complex networks via branch-and-price · Jaewoo Kim, Seungyeop Lee, Jaemyung Ahn et al. · arXiv:2610.26562
A new theory arguing that the structure of networks, like who connects to whom, is actually a fossilized record of strategic choices made by individuals or groups trying to maximize their benefits.
Strategic Origins of Network Structure · Christoph Riedl, Rory Smead · arXiv:2610.24792
A new framework for businesses to decide the "next best action" for customers, like offering a discount or a specific promotion, by learning from past data to maximize long-term value.
Next Best Action: Causal Constrained Offline RL for Business Action Targeting · Surya Ramachandiran, Yiming Li, Juan Mancilla-Caceres et al. · arXiv:2610.26817
This paper models how our individual skills and the widespread adoption of AI might evolve together, showing that too much reliance on AI could lead to a population with lower overall skills or even a divide between highly skilled humans and AI-dependent users.
Modeling the Coevolution of Human Skills and AI Adoption · Giacomo Frigerio, Susanna Caruso, Giacomo Livan et al. · arXiv:2610.24794
A look at how AI systems learn to assign probabilities, especially when guided by larger "teacher" models or pre-existing components. A teacher's underlying bias is easily passed on and often goes undetected by standard fixes, meaning we must check these AI systems against real-world outcomes to ensure they make fair and accurate decisions.
Teacher Supervision, Fixed Representations and Calibration: Exact Risk Calculations for Candidate-Scoring Decision Models · Agus Sudjianto, Denis Burakov · arXiv:2610.26797