Yan Pan

The eruption of civil unrest events often inflicts severe disruption on social stability. Accurate prediction of these events can assist decision-makers in proactively formulating countermeasures, thereby preventing violent incidents and mitigating property damage. Current event prediction methods often focus on predicting a single event element, such as occurrence probability or potential initiators, resulting in less reference information provided to decision-makers and an inability to formulate more specific response measures. To address this limitation, we formulate civil unrest forecasting as a multi-task learning problem and propose a parallel learning framework that jointly predicts social unrest event occurrence and potential initiators. In this framework, an event graph is first encoded by a graph neural module that combines graph attention mechanisms with graph convolutional networks, and the learned event representations are subsequently fed into task-specific predictors for joint forecasting. Extensive comparative experiments conducted across five datasets demonstrate that the proposed framework achieves superior predictive performance and high training efficiency. These findings can support earlier risk identification, improve the allocation of public-safety resources, and provide more informative evidence for mitigating the social disruption caused by civil unrest events.

To address the lack of systematic quantitative evaluation of automated event repositories in forecasting tasks, this study selected Global Database of Events, Language, and Tone (GDELT) and the emerging Political Event Classification, Attributes, and Types (POLECAT) dataset as research subjects, aiming to provide a basis for data source selection through multidimensional comparisons. The primary research question is how to establish a structured, quantitative framework to reliably evaluate these data sources in specific predictive contexts. This study constructed a quantitative framework covering scale, coverage, redundancy, and accuracy, and conducted empirical forecasting tests across multiple cities. The results indicate that while GDELT possesses a large-scale and high media coverage, it performs poorly in terms of redundancy and domain accuracy; although POLECAT is smaller in scale, it exhibits high domain identification accuracy and extremely low redundancy, with its forecast results demonstrating superior precision and false positive control capabilities. The conclusion is that GDELT is suitable for macro-level early warning scenarios requiring high recall, while POLECAT is better suited for tasks requiring high signal-to-noise ratio inputs and specific regional studies; the choice between the two should be based on a trade-off between model requirements and application scenarios.

Persuasion, a vital social skill, influences beliefs, attitudes, and behaviors through conversation. Yet, current dialogue agents either rely on scenario-specific strategies, restricting their cross-context adaptability, or neglect persuasion’s logical structure. They focus on isolated strategy classification, overlooking the significance of fine-grained sequential planning for real-world scenarios. To address these limitations, inspired by basic human mental activities, we present PersuHSG, an adaptive persuasion strategy planning framework. The core idea is to conceptualize persuasion as a tripartite framework comprising cognition, affection, and volition, with each stage represented as a graph layer and principle-based strategies for efficient multi-stage persuasion. Specifically, we first develop PersuInstruct, a fine-tuning dataset to improve dialogue agents’ strategic planning and response generation. Then, we propose a graph-aware planning algorithm for stage-strategy-response reasoning to generate persuasive responses for diverse scenarios. Extensive experiments confirm that PersuHSG significantly enhances the persuasiveness of Large Language Models (LLMs), allows smaller models (e.g., 9B, 13B) to achieve competitive performance, and demonstrates the efficacy of structured strategy planning in improving model efficiency and adaptability.

Instant delivery has become an essential service in daily life, requiring strict delivery timelines. However, traditional delivery methods that employ human couriers struggle to meet the soaring delivery demands due to labor shortages. While researchers have explored alternative solutions using ground vehicles (e.g., crowdsourced taxis) and UnmannedAerial Vehicles (UAVs), their inherent limitations, such as constrained delivery detour for crowdsourced taxis and limited battery capacity of UAVs, greatly constrain their effectiveness. To address these challenges, this paper proposes a novel air-ground delivery paradigm that cooperatively integrates UAVs and crowdsourced taxis. First, UAV stations are strategically deployed based on delivery gaps between the delivery demands and taxis’ delivery capacity, instead of delivery demands only; Then, a predictive UAV repositioning strategy is designed to bridge instantaneously dynamic delivery gaps. Thereafter, a transfer learning-based (TL-based) algorithm that mines the delivery knowledge of human couriers is designed to optimize the cooperative performance. This algorithm extracts behavioral insights from human couriers and transfers them to enhance the delivery capabilities of UAVs and taxis. Finally, parcel assignment is formulated as optimization problems aimed at maximizing total preferences of UAVs and taxis, and maximizing delivery number while minimizing cost, respectively. Evaluations on real-world datasets demonstrate that the proposed method delivers 27.4% more parcels, saves 19.2% delivery cost, and preserves 36.3% more of the travel experience of taxi passengers than the state-of-the-art (SOTA) air-ground cooperative approach for instant delivery.

Unmanned Aerial Vehicles (UAVs) are emerging as a pivotal component in the field of intelligent transportation systems. Leveraging virtual-physical interactions, digital twin technology significantly enhances the adaptability of UAVs in complex traffic environments. However, current approaches still pose three major challenges: contextual adaptability, timely responsiveness, and effective multi-UAV coordination. In this paper, we introduce EnFlexiTwin, a digital twin enhancement assistance platform seamlessly integrated with AdaSor, a lightweight adaptive data selector. EnFlexiTwin automates the construction of incremental learning datasets, enabling real-time enhancements that allow digital twins to adapt to new time series data while preserving historical knowledge. We test EnFlexiTwin on a real-world dataset from low-altitude small-parcel delivery. The results show improved performance and adaptability of digital twins. Furthermore, time-varying simulations on real-world dataset and experiments on a practical air-ground cooperative UAV swarm application highlight that EnFlexiTwin achieves superior enhancements under varying real-time requirements and swarm scale compared to baseline approaches.

This study investigates the development and application of the GDELT (Global Database of Events, Language, and Tone) news database. Through experiments, we conducted a quantitative statistical analysis of the GDELT event database to evaluate its practical characteristics. The results indicate that although the database achieves comprehensive coverage across all countries and regions and includes most major global media outlets, the accuracy rate of its key fields is only approximately 55%, with a data redundancy as high as 20%. Based on these findings, while the GDELT data demonstrates good coverage and data integrity, data correction and deduplication are recommended before its use in research contexts and industrial applications. Subsequently, a survey of the existing literature reveals that current studies using GDELT primarily focused on event-related metrics, such as event quantity, tone, and GoldsteinScale, for application in international relations analysis, crisis event prediction, policy effectiveness testing, and public opinion impact analysis. Nevertheless, news constitutes a fundamental channel of information dissemination in media networks, and the propagation of news events through these networks represents a critical area of study for information recommendation, public opinion guidance, and crisis intervention. Existing research has employed the Event, GKG, and Mentions tables to construct cross-national news flow network models. However, the informational correlations across different data table fields have not been fully leveraged in preliminary data selection, leading to substantial computational overhead. To advance research in this field, this study employs chained list queries on the Event and Mentions tables within GDELT. Using social network analysis, we constructed a media co-occurrence network of event reports, through which core hubs and associative relationships within the event dissemination network are identified.