The University of Texas at Austin,The University of Texas System

Accurately predicting travel mode choice is essential for effective transportation planning, yet traditional statistical and machine learning models are constrained by rigid assumptions, limited contextual reasoning, and reduced transferability. This study explores the potential of Large Language Models (LLMs) as a more flexible and context-aware approach to travel mode choice prediction, enhanced by Retrieval-Augmented Generation (RAG) to ground predictions in empirical data. We develop a modular framework for integrating RAG into LLM-based travel mode choice prediction and evaluate four retrieval strategies: basic RAG, RAG with balanced retrieval, RAG with a cross-encoder for re-ranking, and RAG with balanced retrieval and cross-encoder for re-ranking. These strategies are tested across three LLM architectures (OpenAI GPT-4o, o4-mini, and o3) to examine the interaction between model reasoning capabilities and retrieval methods. Using the 2023 Puget Sound Regional Household Travel Survey data, we conduct a series of experiments to evaluate model performance. The results demonstrate that RAG substantially enhances predictive accuracy across a range of models. Notably, the GPT-4o model combined with balanced retrieval and cross-encoder re-ranking achieves the highest accuracy of 80.8%, exceeding that of conventional statistical and machine learning baselines. Furthermore, LLM–based models exhibit superior zero-shot transfer abilities relative to these baselines. Findings highlight the critical interplay between LLM reasoning capabilities and retrieval strategies, demonstrating the importance of aligning retrieval strategies with model capabilities to maximize the potential of LLM-based travel behavior modeling.

In response to external stimuli, hydrogels undergo large deformation coupled with solvent diffusion. The chemomechanical coupling produces transient, strongly nonlinear responses that can trigger instabilities and bifurcations. Modeling such responses using conventional Finite Element Methods (FEMs) often breaks down due to significant mesh distortion. Here, we present an implicit, mixed material point method (MPM) grounded in non-equilibrium thermodynamics that remains robust under large distortions while consistently coupling deformation and solvent diffusion. Cut-cell ill-conditioning near boundaries is removed using extended B-splines that interpolate boundary degrees of freedom from interior ones via Lagrange polynomials, and subdivision-stabilized interpolation of displacement and chemical potential provides an inf–sup stable discretization. Essential boundary conditions are imposed weakly through a symmetric Nitsche formulation on boundary material points. Together, these improvements yield a symmetric, well-conditioned tangent stiffness suitable for eigenvalue-based stability and bifurcation analysis. We demonstrate the stability and accuracy of the method on benchmark problems, including one-dimensional constrained swelling, swell-induced buckling of a hydrogel column, and bifurcation and surface creasing of a hydrogel with cylindrical pores. The present MPM methodology provides a robust computational foundation for studying extreme, diffusion-coupled deformations in soft-matter applications.

Solar power production will generate tens of millions of tons of end-of-life solar modules by 2050, with glass comprising ∼75% of their mass. This study investigates recycling waste ground solar glass as a supplementary cementitious material (SCM), addressing both PV waste management and the shrinking supply of traditional pozzolans. Solar glasses were predominantly amorphous but had a higher Mg and a lower Fe content than traditional soda-lime glasses. Reactivity of ground solar glasses was governed primarily by fineness and surface pore volume, with the finer solar glass exhibiting the highest dissolution and ASTM C1897 reactivity. Ground solar glass in cement pastes also refined pore structure, consumed portlandite through pozzolanic reaction, and enabled long-term strength recovery. Moreover, trace element leaching remained below regulatory limits, verifying its viability as a safe SCM. Repurposing end-of-life solar glass thereby alleviates waste management burdens and increases SCM supply to construction industry, while fostering sustainable industrial symbiosis.

AI agents may soon become capable of autonomously completing valuable, long-horizon tasks in diverse domains. Current benchmarks either do not measure real-world tasks, or are not sufficiently difficult to meaningfully measure frontier models. To this end, we present Terminal-Bench 2.0: a carefully curated hard benchmark composed of 89 tasks in computer terminal environments inspired by problems from real workflows. Each task features a unique environment, human-written solution, and comprehensive tests for verification. We show that frontier models and agents score less than 65% on the benchmark and conduct an error analysis to identify areas for model and agent improvement. We publish the dataset and evaluation harness to assist developers and researchers in future work at tbench.ai.