arXiv subjects

Hai Yang

Publications and source records attributed to Hai Yang.

At least 19 recordsLinked to original sources

arXiv2026

From Likelihood to Utility: Quality-Axis Diffusion for End-to-End Autonomous Driving

Diffusion models have become a dominant paradigm for end-to-end autonomous driving, yet existing planners organize trajectory generation along the Gaussian-noise axis, treating driving utility only as an auxiliary guidance or selection signal. We instead formulate diffusion along a driving-quality axis. We propose QualiDrive, which constructs scene-specific quality ladders by ranking observed trajectories according to measured driving utility and learns to reverse this ordered degradation through ordered low-to-high-quality refinement. Each reverse step is conditioned on the scene, ego state, current trajectory, and target quality level, progressively refining trajectories toward higher driving quality. On NAVSIM, QualiDrive achieves 94.5 PDMS on v1 and a state-of-the-art 92.74 EPDMS on v2. QualiDrive improves Oracle EPDMS only modestly over endpoint-only quality supervision (98.52 to 98.57), but substantially increases the fractions of candidates above 0.95 and 0.98 EPDMS from 12.94%/9.71% to 21.53%/15.93%, raising selected EPDMS from 92.63 to 92.74. These results suggest that diffusion is not merely a process of denoising, but one of progressive quality-ordered refinement.

cs.RO↗

arXiv2026

EvoAD: Self-Evolving Autonomous Driving via Latent World-Model Guided Post-hoc Reflection

End-to-end autonomous driving systems have achieved rapid progress by learning unified policies that map sensor observations directly to driving trajectories. However, once deployed, these systems typically remain frozen: they may repeat similar decisions in similar scenes, and failures can only be addressed through costly data collection and retraining. This sharply contrasts with human drivers, who improve lifelong through experience, memory, and reflection. In this work, we ask whether autonomous driving systems can achieve similar post-deployment self-evolution to human drivers. We propose EvoAD, a self-evolving autonomous driving framework via latent world-model guided post-hoc reflection. EvoAD stores driving experiences in a bird's-eye-view feature memory pool, retrieves scene-relevant precedents according to geometric and semantic similarity, and uses a latent world model to evaluate the future consequences of candidate trajectories in a compact planning-oriented space. Instead of treating memory as external textual contexts or using world models only for pre-action simulation, EvoAD transforms executed driving outcomes into reusable improvement signals for future decisions. This enables the system to diagnose how prior trajectories could have been improved and to refine behavior in similar future scenarios without retraining the base driving policy. Experiments on NAVSIM v1 and v2 show consistent improvements across reflection rounds and state-of-the-art performance, demonstrating the potential of world models as reflective mechanisms for post-deployment self-evolution in autonomous driving.

cs.RO↗

arXiv2026

EditWM: Event-Decomposed World Modeling with Incremental Correction for End-to-End Autonomous Driving

World models support autonomous driving by predicting the scene evolution associated with candidate trajectories. Driving dynamics differ in predictability, motivating a distinction between regular evolution and event-induced deviations that call for selective correction. We propose EditWM, a world model that decomposes future prediction into normal evolution and event-driven incremental correction in compact visual feature space. A trajectory-conditioned normal predictor provides the base forecast and is then frozen for correction learning. A correction decoder compares this forecast with observation history and planned actions, producing a bounded feature update whose contribution is regulated by a learned gate. The corrected future features condition trajectory scoring through candidate-specific cross-attention, linking world modeling to plan selection. At inference, EditWM uses only past and current observations, ego state, and candidate trajectories. Across all 12,146 NAVSIM navtest scenes, expert-trajectory-conditioned evaluation shows a 5.35\% reduction in future-feature MSE over Normal, with improvements in 83.54\% of scenes. The system achieves 91.05 EPDMS on a 100-point scale using the official EPDMS evaluator. These results demonstrate improved future-feature prediction and competitive trajectory selection when corrected future representations are integrated into planning.

cs.RO↗

arXiv2026

One Click to Leak: Characterizing the Real-World Usage and Threat Impact of MNO-based Single Sign-On Websites

Mobile Network Operator (MNO)-based Single Sign-On (MSSO) is a password-free authentication framework relying on mobile data sessions. Unlike traditional SSO, it shifts the Identity Provider (IdP) to the MNO and the authentication anchor to the Service Provider (SP). MSSO is increasingly deployed and has expanded from mobile apps to websites, yet its web ecosystem and security risks remain largely unexplored. We analyze mainstream MSSO deployments and identify a 3-phase workflow with three trust defects enabling trust hijacking. We further demonstrate One-Click-to-Leak (OCL) attacks, where a single webpage visit can leak sensitive identity information (e.g., phone numbers). With a leading security company, we conduct the first large-scale, longitudinal study of web-based MSSO. We design a hierarchical detection framework using passive DNS correlations and URL reconstruction from search data to identify MSSO-enabled websites. Over one year, we identified 116,852 website URLs across 729 apex domains. Of these URLs, 73.6% exhibit at least one trust defect: 69.4% expose developer credentials, and 27.1% issue high-privilege tokens before user consent, indicating widespread OCL-enabling trust defects. Among the 729 apex domains, 31.8% rely on Resellers, obscuring the downstream SP from the MNO in the analyzed flows. Script analysis identifies 101 websites strongly associated with OCL attack behavior. With our partner, we trace a representative upstream platform subsequently seized by law enforcement and uncover a monetized underground ecosystem. Sanitized backend data shows that it collected 14,100 users' phone numbers within three days and linked them to sensitive information such as browsing activity. Our work provides a comprehensive study of web-based MSSO deployment and security implications. Through responsible disclosure, our work helps secure the mobile authentication ecosystem.

cs.CR↗

arXiv2026

Plug-and-Play Traffic Element Awareness for End-to-End Autonomous Driving

Traffic elements such as traffic lights and road signs play a fundamental role in human driving decisions and should naturally influence end-to-end driving performance. However, existing end-to-end driving research predominantly focuses on dynamic road participants (e.g., vehicles and pedestrians), while the role of traffic elements remains largely unexplored. The community still lacks a systematic study quantifying their impact, largely because public datasets rarely provide structured traffic-element annotations and modern driving systems vary widely in architecture and training paradigm. In this work, we present the first systematic investigation of traffic element awareness for end-to-end autonomous driving. We construct a unified research infrastructure by augmenting multiple public driving datasets with comprehensive traffic-element annotations. To support diverse model families, we adopt a minimal and universal integration design that incorporates traffic-element signals into existing pipelines in a plug-and-play manner with negligible architectural modification. We evaluate this design across modern paradigms, including perception-prediction-planning pipelines, vision-language-action models (VLA), regression-based planners, diffusion-based policies, and trajectory-scoring frameworks, on nuScenes, NAVSIM-v1, NAVSIM-v2, and Bench2Drive. Across all paradigms and datasets, this simple integration consistently improves driving performance, demonstrating that traffic element awareness provides a robust and generalizable signal for end-to-end driving systems. Notably, on the challenging NAVSIM-v2 benchmark, our approach significantly improves state-of-the-art architectures and data pipelines, establishing a new state of the art.

cs.CV↗

arXiv2026

Three-sided mobility-energy market design as a multiperiod stochastic assignment game

As mobility service providers (MSPs) and energy providers (EPs) expand electric vehicle ecosystems, models are needed to understand their interactions within a three-sided market. Existing frameworks often overlook the temporal interdependencies between mobility and charging demands. We address this gap by proposing a bilevel problem as an assignment game overseen by a market regulator. The upper level optimizes service pricing to maximize platform profitability. The lower level models a multi-stakeholder equilibrium using a scalable, link-based Perturbed Utility Route Choice (PURC) framework. The evaluation time frame is divided into discrete intervals, capturing the temporal lag between mobility and charging demand via an empirical affine function. We solve the model by chronologically decomposing the lower level into interacting mobility service and recharge subnetworks. Numerical experiments on the expanded Nguyen-Dupuis network reveal several key insights. First, a critical charging capacity threshold exists; operating below it forces a severe reduction in the deployable fleet and creates localized transit deserts. Second, modeling endogenous operating costs reveals a concave profit trajectory, demonstrating that total profit maximizes at a specific fleet size just before market saturation. Third, optimal dynamic pricing operates within a narrow range, where peak pricing acts as a steady revenue driver and off-peak pricing serves as a highly sensitive operational buffer. These findings provide actionable strategies for coordinating fleet sizing and charging infrastructure deployment.

cs.GT↗

arXiv2026

Deep and diverse population synthesis for multi-person households using generative models with conditional inputs

Traditional methods of population synthesis produce stable and interpretable populations but cannot capture the interrelationships between household- and individual-level attributes. Recent deep learning methods offer this flexibility, yet can overfit high-dimensional attribute relationships without structural guidance and deviate from known structures. We develop a household level synthetic population generation framework that adapts the existing conditional input directed acyclic tabular generative adversarial network, or ciDATGAN, to multi person households. The framework combines household size specific data construction, directed acyclic graphs (DAG) informed dependency regularization, and conditional population inputs as deterministic anchoring to preserve intrahousehold associations. We apply the model to generate an open access synthetic population for New York State. The synthetic population includes nearly 20 million individuals and 7.5 million households in 2021. Validation against withheld benchmarks shows close agreement: joint distribution matching of Public Use Microdata Areas (PUMA), age, race, and disability against the full public use microdata sample (PUMS) yields an R-squared of 0.602 and a Jensen-Shannon distance of 0.280, pairwise Cramer's V differences between generated and benchmark records average below 0.007 at the person level, and classifier two-sample tests on complete household records yield AUC values of 0.527-0.549, close to chance level. The generated households reproduce cross member associations while increasing diversity by 10-17% over the PUMS sample and 13.2% over PopGen alone.

cs.CY↗

arXiv2026

Impacts of radiative cooling on the images of a black hole shadow and extended jets in two-temperature GRMHD simulations

The recent 230 GHz observations from the Event Horizon Telescope collaboration have successfully imaged the supermassive black hole shadow of the M87 galaxy. However, the relatively high radiative efficiency observed in the hot accretion flow suggests that radiative cooling is non-negligible and should be considered when calculating the electron temperature. In this study, we compare accretion models without and with radiative cooling across a range of mass accretion rates, $\dot{M}_{\mathrm{BH}} = (1.0 - 10) \times 10^{-6}\,\dot{M}_{\mathrm{Edd}}$, aiming to assess the impact of cooling on the disk structure, electron temperature distribution (eDF), black hole shadow morphology, broadband spectral energy distributions (SEDs), and flux variability. We performed general relativistic radiative transfer (GRRT) calculations on two-temperature, radiative, general relativistic magnetohydrodynamic (GRMHD) simulations, employing different electron heating prescriptions and nonthermal eDFs, analyzing the radiation transfer due to synchrotron emission at 230 GHz with inclination angle of $163^\circ$. These simulations are targeted toward M87$^{*}$. By comparing density profiles, eDFs, GRRT images, SEDs, and time variability between models, we find that the radiative cooling sharply decreases the electron temperature in the dense inner disk around the equatorial plane ($r\lesssim 10\,r_\mathrm{g}$), while slightly reducing jet sheath temperature. Cooling leads to a dimmer disk, more extended and brighter jets, and reduced total flux. For a given accretion rate, cooling reduces the high-frequency flux. Time variability originates primarily from the midplane in both non-cooling and cooling cases and decreases as accretion rates rise. Although currently below the dynamic range of EHT observations, the features identified in this study could be resolved by next-generation arrays such as the ngEHT.

astro-ph.HE↗

arXiv2026

U3DWind: A Low Altitude Wind Field Dataset and Benchmark for Urban Air Mobility

Urban Air Mobility (UAM) requires reliable assessment of low-altitude wind hazards, because winds, gusts, and building-induced turbulence have been recognized as critical factors affecting vehicle stability, route feasibility, vertiport siting, and airspace management. While wind-tunnel experiments, computational fluid dynamics (CFD), multiscale downscaling, reduced-order models, and UAV planning datasets have advanced wind-aware analysis, public resources for data-driven, city-scale UAM planning remain limited in geographic coverage, scenario diversity, vertical extent, building realism, and task-oriented benchmarking. To address this gap, we introduce U3DWind, a building-resolved low-altitude wind-field dataset generated using our GPU-accelerated Lattice Boltzmann Method--Large-Eddy Simulation (LBM-LES) framework for rapid urban flow simulation. U3DWind covers five megacities in China: Beijing, Shanghai, Guangzhou, Shenzhen, and Hong Kong. It contains 720 simulations, with 16 inflow directions, three reference wind speeds, and three seasonal atmospheric scenarios (annual, summer, and winter) for each city. At a 10 m grid resolution, the dataset provides three-dimensional three-component (3D3C) velocity, turbulent kinetic energy (TKE), flow density, and fluid--solid masks. To support operationally relevant evaluation, we further define five baseline tasks: wind-field prediction, sparse-sensor wind-field reconstruction, site wind-exposure ranking, airworthiness wind-compliance risk scoring, and noise propagation modeling. As a multi-city, building-resolved 3D urban wind-field dataset, U3DWind enables systematic evaluation of wind-induced impacts in low-altitude traffic scenarios and provides an open benchmark for urban airspace management and data-driven high-fidelity urban flow simulation.

physics.flu-dyn↗

arXiv2026

Distance-based subsidy rate design to incentivize ride-hail access to advanced air mobility hubs

The success of advanced air mobility (AAM) operations is largely contingent on its effective integration with other ground transport modes. Under many use cases, AAM operators have to work with ride-hailing operators to create a seamless air taxi travel experience with adequate first and last-mile access. In investigating this multimodal coalition, this study proposes a distance-based subsidy rate design for AAM operators to incentivize ride-hail access to AAM hubs, incorporating air mobility operators' profitability considerations and travelers' route choices jointly. Using New York City (NYC) airport access as a case study, this study integrates high-volume for-hire vehicle (HVFHV) data from NYC taxi zones to consider real-world spatial demand distributions while considering passenger groups with different values of time (VOT) to derive insights on distinctive customer bases. Overall, the results show that AAM operators would need to subsidize the ride-hailing operators on vertiport access trips when air taxi operating costs exceed $12/mi. The analysis of ridership at AAM hubs indicates that ridership and profit contributions differ across different candidate vertiports in Manhattan, reflecting spatial demand heterogeneity. Additionally, having the airport access system in place, the taxi zones that generate the highest passenger demand to all three major NYC airports are identified under lower air taxi fare scenarios. These findings highlight how a distance-based subsidy rate design is beneficial in facilitating better access to vertiports and to foster high air taxi ridership with optimal AAM fare levels.

cs.CY↗

arXiv2026

LLM-ODDR: A Large Language Model Framework for Joint Order Dispatching and Driver Repositioning

Ride-hailing platforms face significant challenges in optimizing order dispatching and driver repositioning operations in dynamic urban environments. Traditional approaches based on combinatorial optimization, rule-based heuristics, and reinforcement learning often overlook driver income fairness, interpretability, and adaptability to real-world dynamics. To address these gaps, we propose LLM-ODDR, a novel framework leveraging Large Language Models (LLMs) for joint Order Dispatching and Driver Repositioning (ODDR) in ride-hailing services. LLM-ODDR framework comprises three key components: (1) Multi-objective-guided Order Value Refinement, which evaluates orders by considering multiple objectives to determine their overall value; (2) Fairness-aware Order Dispatching, which balances platform revenue with driver income fairness; and (3) Spatiotemporal Demand-Aware Driver Repositioning, which optimizes idle vehicle placement based on historical patterns and projected supply. We also develop JointDR-GPT, a fine-tuned model optimized for ODDR tasks with domain knowledge. Extensive experiments on real-world datasets from Manhattan taxi operations demonstrate that our framework significantly outperforms traditional methods in terms of effectiveness, adaptability to anomalous conditions, and decision interpretability. To our knowledge, this is the first exploration of LLMs as decision-making agents in ride-hailing ODDR tasks, establishing foundational insights for integrating advanced language models within intelligent transportation systems. While the current framework incurs higher computational costs than traditional methods, we show that parallel decomposition and model distillation can reduce latency to production-viable levels for deployment.

cs.LG↗

arXiv2026

A Jet from a Nearly Dormant Black Hole

Most galaxies host supermassive black holes (SMBHs) that remain weakly accreting or dormant for much of their lifetimes. At the lowest accretion rates, these systems may represent the transition between active nuclei and dormant black holes, but whether they can still launch collimated jets remains unclear. The nuclei in our Galaxy (\sgra) and M31 are key examples of this regime, although no clear jet structure has yet been detected in either source. Here we report multi-frequency very long baseline interferometric observations of \Msixty\ (NGC~4649), a nearby elliptical galaxy hosting a nearly dormant SMBH with an Eddington ratio of $\sim10^{-8}$. We detect a compact two-sided jet with an unusually steep synchrotron spectrum, demonstrating that collimated outflows can persist even under nearly dormant accretion conditions. The apparent radio core exhibits an unprecedentedly steep frequency-dependent position shift toward the SMBH, locating the central engine only $\sim57\,μ$as, corresponding to a projected distance of $\sim10$ Schwarzschild radii, upstream of the 8.37-GHz core. The observed jet morphology and steep core-shift behaviour are reproduced by general relativistic magnetohydrodynamic and radiative-transfer simulations, indicating a magnetically dominated, non-equipartition jet-launching region that departs from the standard conical equipartition picture. These results provide direct observational evidence that jet production can survive near the dormant SMBHs and establish \Msixty\ as a unique laboratory for probing jet formation on event-horizon scales in the lowest-accretion SMBH regime.

astro-ph.HE↗

arXiv2026

Think Before You Drive: World Model-Inspired Multimodal Grounding for Autonomous Vehicles

Interpreting natural-language commands to localize target objects is critical for autonomous driving (AD). Existing visual grounding (VG) methods for autonomous vehicles (AVs) typically struggle with ambiguous, context-dependent instructions, as they lack reasoning over 3D spatial relations and anticipated scene evolution. Grounded in the principles of world models, we propose ThinkDeeper, a framework that reasons about future spatial states before making grounding decisions. At its core is a Spatial-Aware World Model (SA-WM) that learns to reason ahead by distilling the current scene into a command-aware latent state and rolling out a sequence of future latent states, providing forward-looking cues for disambiguation. Complementing this, a hypergraph-guided decoder then hierarchically fuses these states with the multimodal input, capturing higher-order spatial dependencies for robust localization. In addition, we present DrivePilot, a multi-source VG dataset in AD, featuring semantic annotations generated by a Retrieval-Augmented Generation (RAG) and Chain-of-Thought (CoT)-prompted LLM pipeline. Extensive evaluations on six benchmarks, ThinkDeeper ranks #1 on the Talk2Car leaderboard and surpasses state-of-the-art baselines on DrivePilot, MoCAD, and RefCOCO/+/g benchmarks. Notably, it shows strong robustness and efficiency in challenging scenes (long-text, multi-agent, ambiguity) and retains superior performance even when trained on 50% of the data.

cs.CV↗

arXiv2026

Do Open-Loop Metrics Predict Closed-Loop Driving? A Cross-Benchmark Correlation Study of NAVSIM and Bench2Drive

Open-loop evaluation offers fast, reproducible assessment of autonomous driving planners, but its ability to predict real closed-loop driving performance remains questionable. Prior work has shown that traditional open-loop metrics such as Average Displacement Error (ADE) and Final Displacement Error (FDE) exhibit no reliable correlation with closed-loop Driving Score. In this paper, we ask whether the more recent, safety-aware open-loop metrics introduced by NAVSIM~v2 can bridge this gap. By systematically cross-referencing published results from 15 state-of-the-art methods across NAVSIM (open-loop) and Bench2Drive (closed-loop), we compile a paired dataset of open-loop sub-metrics and closed-loop performance, yielding 8 methods with complete paired data. Our analysis reveals three key findings: (1) the aggregate NAVSIM PDM Score shows a strong positive but non-monotonic correlation with Bench2Drive Driving Score, with clear ranking inversions; (2) among individual NAVSIM sub-metrics, Ego Progress (EP) is the strongest single predictor of closed-loop success, substantially exceeding the safety-critical collision metric NC; (3) the safety-progress trade-off manifests differently in open-loop and closed-loop: methods that maximize safety at the expense of progress rank highly in NAVSIM but underperform in closed-loop due to timeout and slow-driving penalties. We further demonstrate that a much simpler 3-metric formula matches the predictive power of the full 5-metric PDMS at the same Spearman $ρ{=}0.90$ on our paired sample of $n{=}8$ methods, suggesting that within current state-of-the-art methods -- where TTC and Comfort approach saturation -- these two sub-metrics add little marginal information for closed-loop ranking. Additionally, we identify the snowball effect -- where small open-loop deviations compound into closed-loop failures -- as a candidate mechanism for the residual gap.

cs.RO↗

arXiv2026

Welfare, sustainability, and equity evaluation of the New York City Interborough Express using spatially heterogeneous mode choice models

The Metropolitan Transit Authority (MTA) proposed building a new light rail route called the Interborough Express (IBX) to provide a direct, fast transit linkage between Queens and Brooklyn. An open-access synthetic citywide trip agenda dataset and a block-group-level mode choice model are used to assess the potential impact IBX could bring to New York City (NYC). IBX could save 28.1 minutes to potential riders across the city. For travelers either going to or departing from areas close to IBX, the average time saving is projected to be 29.7 minutes. IBX is projected to have more than 272 thousand daily ridership after its completion (81% higher than reported in the official IBX proposal). Among those riders, more than 58 thousand people (21.4%) would come from low-income households while 185 thousand people (68.2%) would start or end along the IBX corridor. The addition of IBX would attract more than 40 thousand additional daily trips to transit mode, among which more than 16 thousand would be switched from using private vehicles, reducing potential greenhouse gas (GHG) emissions by 30.63 metric tons per day. IBX can also bring significant consumer surplus benefits to the communities, which are estimated to be $0.89 USD per trip. However, the service does not appear to significantly reduce the proportion of travelers whose consumer surpluses fall below 10% of the population average (already quite low).

cs.CY↗

arXiv2026

SAIL: Scene-aware Adaptive Iterative Learning for Long-Tail Trajectory Prediction in Autonomous Vehicles

Autonomous vehicles (AVs) rely on accurate trajectory prediction for safe navigation in diverse traffic environments, yet existing models struggle with long-tail scenarios-rare but safety-critical events characterized by abrupt maneuvers, high collision risks, and complex interactions. These challenges stem from data imbalance, inadequate definitions of long-tail trajectories, and suboptimal learning strategies that prioritize common behaviors over infrequent ones. To address this, we propose SAIL, a novel framework that systematically tackles the long-tail problem by first defining and modeling trajectories across three key attribute dimensions: prediction error, collision risk, and state complexity. Our approach then synergizes an attribute-guided augmentation and feature extraction process with a highly adaptive contrastive learning strategy. This strategy employs a continuous cosine momentum schedule, similarity-weighted hard-negative mining, and a dynamic pseudo-labeling mechanism based on evolving feature clustering. Furthermore, it incorporates a focusing mechanism to intensify learning on hard-positive samples within each identified class. This comprehensive design enables SAIL to excel at identifying and forecasting diverse and challenging long-tail events. Extensive evaluations on the nuScenes and ETH/UCY datasets demonstrate SAIL's superior performance, achieving up to 28.8% reduction in prediction error on the hardest 1% of long-tail samples compared to state-of-the-art baselines, while maintaining competitive accuracy across all scenarios. This framework advances reliable AV trajectory prediction in real-world, mixed-autonomy settings.

cs.ET↗

arXiv2026

Language-Grounded Multi-Agent Planning for Personalized and Fair Participatory Urban Sensing

Participatory urban sensing leverages human mobility for large-scale urban data collection, yet existing methods typically rely on centralized optimization and assume homogeneous participants, resulting in rigid assignments that overlook personal preferences and heterogeneous urban contexts. We propose MAPUS, an LLM-based multi-agent framework for personalized and fair participatory urban sensing. In our framework, participants are modeled as autonomous agents with individual profiles and schedules, while a coordinator agent performs fairness-aware selection and refines sensing routes through language-based negotiation. Experiments on real-world datasets show that MAPUS achieves competitive sensing coverage while substantially improving participant satisfaction and fairness, promoting more human-centric and sustainable urban sensing systems.

cs.AI↗

arXiv2026

Can Large Multimodal Models Inspect Buildings? A Hierarchical Benchmark for Structural Pathology Reasoning

Automated building facade inspection is a critical component of urban resilience and smart city maintenance. Traditionally, this field has relied on specialized discriminative models (e.g., YOLO, Mask R-CNN) that excel at pixel-level localization but are constrained to passive perception and worse generization without the visual understandng to interpret structural topology. Large Multimodal Models (LMMs) promise a paradigm shift toward active reasoning, yet their application in such high-stakes engineering domains lacks rigorous evaluation standards. To bridge this gap, we introduce a human-in-the-loop semi-automated annotation framework, leveraging expert-proposal verification to unify 12 fragmented datasets into a standardized, hierarchical ontology. Building on this foundation, we present \textit{DefectBench}, the first multi-dimensional benchmark designed to interrogate LMMs beyond basic semantic recognition. \textit{DefectBench} evaluates 18 state-of-the-art (SOTA) LMMs across three escalating cognitive dimensions: Semantic Perception, Spatial Localization, and Generative Geometry Segmentation. Extensive experiments reveal that while current LMMs demonstrate exceptional topological awareness and semantic understanding (effectively diagnosing "what" and "how"), they exhibit significant deficiencies in metric localization precision ("where"). Crucially, however, we validate the viability of zero-shot generative segmentation, showing that general-purpose foundation models can rival specialized supervised networks without domain-specific training. This work provides both a rigorous benchmarking standard and a high-quality open-source database, establishing a new baseline for the advancement of autonomous AI agents in civil engineering.

cs.CV↗