arXiv subjects

Ziqi Song

Publications and source records attributed to Ziqi Song.

6 recordsLinked to original sources

arXiv2026

HERMES: a multi-agent framework for structured knowledge extraction from ultra-long documents in geoscience

Authoritative scientific knowledge in geoscience remains largely trapped in legacy monographs and historical literature, where unstructured text and complex layouts hinder computational access. We introduce HERMES, a scalable multi-agent framework that extracts structured data from ultra-long scientific documents. Using a coordinating large language model, HERMES integrates domain constraints, validation rules and evidence tracing within a unified document-level extraction process that incorporates parsed text, tables, figures and captions. Applied to the 55-volume Treatise on Invertebrate Paleontology, the system produced a structured database of 32,277 fossil taxonomic entities and 451,878 attributes, released online at https://treatise.geolex.org. Extraction performance remained stable across fossil groups (average F1 scores of approximately 0.90 for entities and 0.91 for attributes), improving per-volume efficiency approximately sixfold relative to the tested fully manual baseline. Evaluation in palaeomagnetism and geochemistry, conducted without additional model training, demonstrated transfer across distinct geoscience domains. This work provides a practical pathway to transform historical scientific literature into FAIR-oriented structured data, offering a sustainable infrastructure for data-intensive disciplines and large-scale knowledge integration.

cs.CL↗

arXiv2025

School Transport Electrification -- Adoption, Strategies, Methods and Policy: A Comprehensive Review

The move towards electric school buses (ESBs) marks a critical step in creating a healthier and more sustainable future for students. To meet the ambitious goal of zero-emission school buses by 2035,this review focuses on the need assessment, practices, gaps, challenges, and way forward. We conducted a comprehensive assessment of more than 100 relevant sources, resulting in a final investigation. In-depth, systematic, and qualitative content analysis with SWOT analysis produced critical insights into school transport electrification. The results showed that 1.8% of the total buses in the US have already been converted to electric, where California alone owning 29% of the buses. Subsidies from various agencies and programs have contributed to the rapid growth of electrification. However, challenges in cost, technology, and policies must be mitigated through innovation and stakeholder partnerships. Policy support is boosting subsidies, industry investment and market readiness. Equitable policy is important to support underserved and disadvantaged populations, which can be addressed through four key dimensions of equity: procedural, recognition, distributive, and reparative equity. Furthermore, the traditional bus deployment model is still the most common, whereas Transportation-as-a-Service (TaaS) is an innovative ESB deployment model with the potential to accelerate ESB adoption by integrating vehicle-to-grid. SWOT analysis indicated that the achievement of the zero-emission goal, autonomous driving, and repowered vehicle technology are the greatest opportunities. Dynamic electrification strategies, V2G technology and system resiliency are yet to be discovered, which could be crucial for mass electrification.

physics.soc-ph↗

arXiv2025

Understanding Mode Choice Behavior of People with Disabilities: A Case Study in Utah

Despite the growing recognition of the importance of inclusive transportation policies nationwide, there is still a gap, as the existing transportation models often fail to capture the unique travel behavior of people with disabilities. This research study focuses on understanding the mode choice behavior of individuals with travel-limited disabilities and comparing the group with no such disability. The study identified key factors influencing mode preferences for both groups by utilizing Utah's household travel survey, simulation algorithm and Multinomial Logit model. Explanatory variables include household and socio-demographic attributes, personal, trip characteristics, and built environment variables. The analysis revealed intriguing trends, including a shift towards carpooling among disabled individuals. People with disabilities placed less emphasis on travel time saving. A lower value of travel time for people with disabilities is potentially due to factors like part-time work, reduced transit fare, and no or shared cost for carpooling. Despite a 50% fare reduction for the disabled group, transit accessibility remains a significant barrier in their choice of Transit mode. In downtown areas, people with no disability were found to choose transit compared to driving, whereas disabled people preferred carpooling. Travelers with no driving licenses and disabled people who use transit daily showed complex travel patterns among multiple modes. The study emphasizes the need for accessible and inclusive transportation options, such as improved public transit services, shorter first and last miles in transit, and better connectivity for non-motorized modes, to cater to the unique needs of disabled travelers. The findings of this study have significant policy implications such as an inclusive mode choice modeling framework for creating a more sustainable and inclusive transportation system.

physics.soc-ph↗

arXiv2025

Evolving School Transport Electrification: Integrated Dynamic Route Optimization and Partial Charging for Mixed Fleets

School bus transportation, the largest fleet size for public transportation in the US, plays a significant role in sustainability through transport decarbonization. Thus, effective planning of electric school bus routes and recharge schedules is vital. This study proposes a novel approach that simultaneously addresses electric school bus dynamic routing and partial charge scheduling, considering practical scenarios such as varying student demands, bus capacities, maximum ride time, stop time window, and fleet mixes. The model incorporates constraints like bell time tolerance and battery capacity and charging infrastructure candidate location, making it robust for school bus electrification. A linearized Mixed Integer Programming (MIP) model for homogeneous and heterogeneous fleets with full and partial recharging strategies is formulated. The proposed objective function for nonlinear and linear models is executed and compared for computational effectiveness. The model is tested on various sizes of school networks using modified benchmark instances, and a real-world case study demonstrates the benefits of electrified school transportation. The results show that employing heterogeneous fleets can lead to cost savings, reduced routing distance, and travel time for both the tested networks and the case study. Sensitivity analyses highlight the trade-offs between battery size and total cost. Furthermore, the benefits of partial charging and optimum riding time for school bus routes are suggested. The proposed optimization approach can achieve significant reductions in travel distance, up to 56.4% compared to the current situation and fleet size, supporting the case for school transport electrification. Potential additional investment subsidies from federal and state governments are added benefits for accelerated school bus electrification.

math.OC↗

arXiv2025

Table2LaTeX-RL: High-Fidelity LaTeX Code Generation from Table Images via Reinforced Multimodal Language Models

In this work, we address the task of table image to LaTeX code generation, with the goal of automating the reconstruction of high-quality, publication-ready tables from visual inputs. A central challenge of this task lies in accurately handling complex tables -- those with large sizes, deeply nested structures, and semantically rich or irregular cell content -- where existing methods often fail. We begin with a comprehensive analysis, identifying key challenges and highlighting the limitations of current evaluation protocols. To overcome these issues, we propose a reinforced multimodal large language model (MLLM) framework, where a pre-trained MLLM is fine-tuned on a large-scale table-to-LaTeX dataset. To further improve generation quality, we introduce a dual-reward reinforcement learning strategy based on Group Relative Policy Optimization (GRPO). Unlike standard approaches that optimize purely over text outputs, our method incorporates both a structure-level reward on LaTeX code and a visual fidelity reward computed from rendered outputs, enabling direct optimization of the visual output quality. We adopt a hybrid evaluation protocol combining TEDS-Structure and CW-SSIM, and show that our method achieves state-of-the-art performance, particularly on structurally complex tables, demonstrating the effectiveness and robustness of our approach.

cs.AI↗

arXiv2022

Bond-Selective Intensity Diffraction Tomography

Recovering molecular information remains a grand challenge in the widely used holographic and computational imaging technologies. To address this challenge, we developed a computational mid-infrared photothermal microscope, termed Bond-selective Intensity Diffraction Tomography (BS-IDT). Based on a low-cost brightfield microscope with an add-on pulsed light source, BS-IDT recovers both infrared spectra and bond-selective 3D refractive index maps from intensity-only measurements. High-fidelity infrared fingerprint spectra extraction is validated. Volumetric chemical imaging of biological cells is demonstrated at a speed of ~20 seconds per volume, with a lateral and axial resolution of ~350 nm and ~1.1 micron, respectively. BS-IDT's application potential is investigated by chemically quantifying lipids stored in cancer cells and volumetric chemical imaging on Caenorhabditis elegans with a large field of view (~100 micron X 100 micron).

physics.optics↗