2026-10-05

7 arXiv categories· 96 journals· 4665 candidates — 4 worth your time· 52 without an open abstract

Headline

Mobility-based policy evaluation, trajectory diversity loss, and LLM-enhanced bike-demand forecasting

Today, together

tag shift
no tag ran above its 30-day average
on the same shoulders
no foundation work was cited by four of the day's papers counted over the 48 of this day's 56 named papers whose reference lists we have — standing on the same older work is not the same as being about the same thing
first appearance
no institution here was new to the last 115 days none of them on a paper we published in the 115 days before this issue

Two counts, not one counted twice: every row here but “on the same shoulders” counts the papers published above, and that one counts every paper this day named — those and the ones under Also published today, whose abstracts we could not read.

An agent-based simulation model was developed to project changes in trajectory diversity from 2026 to 2036 across China, Russia, Japan, the United Kingdom, and the United States. Another study evaluated three transportation interventions in Boston and New York City using both traditional geographic distance and mobility-network exposure to show how human mobility shapes the spread of urban policy effects. Finally, researchers developed Wellbeing-Former, a reasoning model designed to measure flood-related wellbeing across Ireland using public discourse.

preprint

Bijin Joseph, Hamish Gibbs, Takahiro Yabe, et al.
arXiv physics.soc-ph

Urban Transport and Accessibility · Human Mobility and Location-Based Analysis · Transportation Planning and Optimization · causal inference · Boston · New York City

Human mobility shapes how urban policy effects spread across cities

This study evaluates three transportation interventions in Boston and New York City using both traditional geographic distance and mobility-network exposure, leveraging large-scale mobility data. In Boston, a free-fare transit program was found to increase overall activity by roughly 16%, particularly in lower-income areas, while Open Streets programs boosted visits to local food businesses. For Manhattan's congestion pricing, a mobility-based analysis revealed no net reduction in activity but a shift away from shopping and grocery trips, contrasting with geographic models that incorrectly predicted a general decline.

Why it matters — It demonstrates that evaluating urban policies based on spatial proximity rather than actual human mobility networks systematically distorts or misses their behavioral impacts. By analyzing 138 metropolitan areas across eight countries, the study establishes that geographic and mobility distances diverge locally in every city, proving that mobility-based exposure is a universally necessary framework for policy evaluation.

arXiv · pdf

preprint

Daneul Kim, Yuyeong Kim
arXiv cs.CY

agent-based model · China · United Kingdom · United States

Conditions for Social Trajectory Collapse: Agent-Based Simulation of Time-Geographic Trajectory Distributions

An agent-based simulation model was developed to project changes in trajectory diversity—the variety of recurring activity orientations toward neighborhoods and urban centers—from 2026 to 2036 across five countries: China, Russia, Japan, the United Kingdom, and the United States. The model, initialized with 2010 population shares, projects a 52.6% to 79.0% decline in trajectory diversity over the decade, alongside rising cost burdens in all cases and falling welfare in Russia and Japan.

Why it matters — The study establishes a new framework for evaluating regional development policies by demonstrating that spatial concentration leads to a severe contraction in the variety of daily human activity patterns, a dimension of urban life that is overlooked by traditional economic metrics.

Caveat: The findings rely entirely on simulated agent behaviors and projected population distributions rather than observed empirical trajectories.

arXiv · pdf

preprint

Róisín Luo, Karyn Morrissey
arXiv cs.CY

Flood Risk Assessment and Management · Climate Change Communication and Perception · Disaster Management and Resilience

A population-level assessment framework for flood-related wellbeing from public discourse in Ireland

The researchers developed Wellbeing-Former, a reasoning model designed to measure flood-related wellbeing across three dimensions: distress, functional disruption, and institutional alienation. The model was applied to a dataset of approximately 224,000 social media posts from Ireland containing flood-related terms, spanning the years 2012 to 2016.

Why it matters — It establishes a method to continuously monitor the psychological and social impacts of climate hazards at a population scale, capturing subtle human responses like institutional alienation that traditional economic damage assessments overlook.

Caveat: The study relies on social media posts from a single platform as a proxy for public discourse, which may introduce demographic biases in who is represented.

arXiv · pdf

preprint

Xuxilu Zhang, Francesc Soriguera
arXiv cs.LG

large language model · Barcelona

LLM-enhanced spatio-temporal learning for grid-level docked bike sharing demand prediction

The proposed forecasting framework, ZI-ASTGCN, uses a Large Language Model to convert unstructured urban text like event schedules and news into quantified spatial-temporal perturbation fields. The model extracts intensity, spatial reach, and temporal lag parameters to construct Gaussian decay fields, which are integrated with a dual-branch output head and a multi-task zero-inflated loss to handle sparse grid-level data. Evaluated on the Barcelona Bicing bike-sharing dataset, the model outperformed established neural baselines, showing significant accuracy improvements during high-demand periods.

Why it matters — It establishes a systematic method to inject unstructured, real-time textual information about urban events directly into numerical spatial-temporal deep learning models, allowing forecasting systems to adapt to sudden travel demand shocks that historical flow data alone cannot predict.

Caveat: The framework's performance and validation are demonstrated using a single bike-sharing dataset in Barcelona.

arXiv · pdf

Also published today

These appeared today in journals we track. Their abstracts are not openly available, so we cannot summarise them.