arXiv subjects

Gui Zhang

Publications and source records attributed to Gui Zhang.

4 recordsLinked to original sources

arXiv2026

Coordinated incentives in AI-generated misinformation governance

With the rapid diffusion of AI-generated content, AI-driven misinformation is becoming increasingly pervasive and difficult to govern, undermining information credibility and social trust. This study models the strategic interdependence among a government regulator, an AI enterprise, and users through a three-party evolutionary game that incorporates heterogeneous rewards and punishments. From the resulting replicator equations, we characterize the evolutionary stability of competing governance and production strategies. The analysis indicates that neither unilateral regulation nor market incentives alone can effectively curb misinformation. Instead, an evolutionarily stable regime of real-information production arises only when regulatory rewards and punishment intensity, enterprise reputation loss, and user adoption incentives collectively surpass critical thresholds. The findings highlight the need for coordinated and adaptive policy mixes that align regulatory instruments with enterprise behavior and user uptake while managing governance costs.

physics.soc-ph↗

arXiv2025

Investigating training objective for flow matching-based speech enhancement

Speech enhancement(SE) aims to recover clean speech from noisy recordings. Although generative approaches such as score matching and Schrodinger bridge have shown strong effectiveness, they are often computationally expensive. Flow matching offers a more efficient alternative by directly learning a velocity field that maps noise to data. In this work, we present a systematic study of flow matching for SE under three training objectives: velocity prediction, $x_1$ prediction, and preconditioned $x_1$ prediction. We analyze their impact on training dynamics and overall performance. Moreover, by introducing perceptual(PESQ) and signal-based(SI-SDR) objectives, we further enhance convergence efficiency and speech quality, yielding substantial improvements across evaluation metrics.

cs.SD↗

arXiv2025

Spatial public goods games with queueing and reputation

In real-world social and economic systems, the provisioning of public goods generally entails continuous interactions among individuals, with decisions to cooperate or defect being influenced by dynamic factors such as timing, resource availability, and the duration of engagement. However, the traditional public goods game ignores the asynchrony of the strategy adopted by players in the game. To address this problem, we propose a spatial public goods game that integrates an M/M/1 queueing system to simulate the dynamic flow of player interactions. We use a birth-death process to characterize the stochastic dynamics of this queueing system, with players arriving following a Poisson process and service times being exponentially distributed under a first-come-first-served basis with finite queue capacity. We also incorporate reputation so that players who have cooperated in the past are more likely to be chosen for future interactions. Our research shows that a high arrival rate, low service rate, and the reputation mechanism jointly facilitate the emergence of cooperative individuals in the network, which thus provides an interesting and new perspective for the provisioning of public goods.

cs.SI↗

arXiv2024

The evolution of cooperation in spatial public goods game with tolerant punishment based on reputation threshold

Reputation and punishment are significant guidelines for regulating individual behavior in human society, and those with a good reputation are more likely to be imitated by others. In addition, society imposes varying degrees of punishment for behaviors that harm the interests of groups with different reputations. However, conventional pairwise interaction rules and the punishment mechanism overlook this aspect. Building on this observation, this paper enhances a spatial public goods game in two key ways: 1) We set a reputation threshold and use punishment to regulate the defection behavior of players in low-reputation groups while allowing defection behavior in high-reputation game groups. 2) Differently from pairwise interaction rules, we combine reputation and payoff as the fitness of individuals to ensure that players with both high payoff and reputation have a higher chance of being imitated. Through simulations, we find that a higher reputation threshold, combined with a stringent punishment environment, can substantially enhance the level of cooperation within the population. This mechanism provides deeper insight into the widespread phenomenon of cooperation that emerges among individuals.

cs.SI↗