Higher-order network adaptivity: co-evolution of higher-order structure and spreading dynamics

Longzhao Liu 1,3,5,6,7, Hongwei Zheng 8, Zhihao Han 1,3, Xin Wang 1,3,5,6,7, Shaoting Tang 1,2,3,4,5,6,7 · 2025

Read Higher-order network adaptivity: co-evolution of higher-order structure and spreading dynamics by Longzhao Liu 1,3,5,6,7, Hongwei Zheng 8, Zhihao Han 1,3, Xin Wang 1,3,5,6,7, Shaoting Tang 1,2,3,4,5,6,7 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-15445.txt.

Introduction

Longzhao Liu and four co-authors let group links in a network reshape as a spread runs through them, in a paper published in 2025. Real spread rides group chats, not just pairs. Earlier adaptive work stayed with pair links. Their build lets the group web itself shift with the spread, and they work out its starting thresholds and final sizes.

Pair-like adaptation and group adaptation both raise the bar to start a spread. Past that, they split: group adaptation shrinks or wipes out the two-state zone where a spread can hang or die. It also turns sharp phase jumps into smooth ones. Tests on synthetic and real group webs support the mathematics. The authors say group adaptation is its own phenomenon, not a pair echo.

Catalog record

CREATORLongzhao Liu 1,3,5,6,7, Hongwei Zheng 8, Zhihao Han 1,3, Xin Wang 1,3,5,6,7, Shaoting Tang 1,2,3,4,5,6,7
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE23 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-15445

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Higher-order network adaptivity: co-evolution of higher-order structure and spreading dynamics Longzhao Liu 1,3,5,6,7, Hongwei Zheng 8, Zhihao Han 1,3, Xin Wang 1,3,5,6,7, Shaoting Tang 1,2,3,4,5,6,7 Abstract The co-evolution of structure and dynamics, known as adaptivity, is a fundamental property in various systems and drives diverse emergent behaviors. However, the adaptivity in previous works is primarily stemmed from pairwise situations, while is insufficient to capture ubiquitous higher-order characteristics of real systems. Here, we introduce higher-order network adaptivity to model the co-evolution of higher-order structure and spreading dynamics, and theoretically analyze the thresholds and spreading sizes. Results demonstrate that both pairwise-like and higher-order adaptivity consistently increase spreading thresholds, but surprisingly produce completely opposing qualitative effects. Specifically, contrary to pairwise-like adaptivity, higher-order adaptivity not only reduces or even eliminates the bistable region, but also leads to shifts of phase transitions from discontinuous to continuous. These findings are validated on both synthetic and real hypergraphs. Our work introduces an idea of higher-order adaptivity and highlights its fundamental differences from pairwise-like adaptivity, advancing further researches of adaptive higher-order systems. Complex systems, ranging from social to natural domains, can typically be modeled as networks and the dynamics on top of them, which jointly govern the emergent behaviors Boccaletti et al. 2006 ; Dorogovtsev et al. 2008 ; Barrat et al. 2008 ; Zhu et al. 2024 . In paricular, networks usually co-evolve with the dynamics of nodes’ states in real-world systems Berner et al. 2023 ; Gross and Blasius 2008 ; Liu et al. 2020 . For instance, in the spreading of infectious diseases, healthy individuals may avoid contacts with infected ones, causing changes of network structure, which in turn affects the spreading dynamics. The mutual influence between structure and dynamics, known as adaptivity, is identified as core mechanism in various systems and drives rich phenomena, such as bistability Gross et al. 2006 , echo chamber Wang et al. 2020 and cooperation Fu et al. 2008 . However, these works mainly concentrate on pairwise situations which assume all interactions between units are dyadic. Recently, higher-order interactions among multiple units have been empirically observed across diverse systems, and exhibit fundamental differences from pairwise interactions at multiple levels Battiston et al. 2020 ; Santoro et al. 2023 ; Grilli et al. 2017 ; Burgio et al. 2024 ; Lambiotte et al. 2019 . Structurally, such interactions require characterizations using higher-order structure such as hyperedges and simplicial complexes, which can not be decomposed into pairwise links Shi and Chen 2022 ; Boccaletti et al. 2023 . Dynamically, higher-order interactions are typically manifested as nonlinear dynamical mechanisms that lead to entirely new emergent behaviors beyond pairwise interactions in various real-world systems Ferraz de Arruda et al. 2024 ; Battiston et al. 2021 ; Majhi et al. 2022 ; Malizia et al. 2025 ; Neuhäuser et al. 2020 . For instance, higher-order spreading mechanism can induce the emergence of bistability and discontinuous phase transitions Iacopini et al. 2019 . Higher-order games may result in explosive transition to cooperation Civilini et al. 2024 . In addition, recent studies provide frameworks of temporal higher-order networks, and reveal its unique topological properties and impacts on dynamical behaviors like percolation time and contagion thresholds Di Gaetano et al. 2024 ; Petri and Barrat 2018 ; Han et al. 2024 ; Iacopini et al. 2024 . Despite the progress, most works study higher-order network and dynamics separately, whereas real-world systems usually exhibit co-evolution of higher-order structure and dynamics, namely higher-order adaptivity. Take the spreading of infectious diseases as a paradigmatic example St-Onge et al. 2021 ; Wang et al. 2024 ; Chen et al. 2024 . Individuals tend to leave groups with infection risks Gross et al. 2006 ; Marceau et al. 2010 , and the possibility of group breakdown may nonlinearly increases as the number of infected members grows Burgio et al. 2025 . For instance, during Covid-19, individuals in low-risk groups with few infected members adopt minimal protective measures, while high-risk areas implement strict lockdown Lai et al. 2020 ; Perra 2021 ; Flaxman et al. 2020 . This reflects the evolution of higher-order structure induced by the number of infected individuals, which in turn affects the epidemic spreading, forming a co-evolutionary process. Nevertheless, the basic rule of higher-order adaptivity and its complex impacts on dynamical behaviors remain unclear. In this paper, we propose a higher-order adaptive spreading model to describe the co-evolution of higher-order networks and spreading dynamics. We provide a theoretical framework that could well predict outbreak sizes and thresholds. Results on both synthetic and real hypergraphs exhibit rich and robust impacts of higher-order adaptivity. Firstly, it can increase the epidemic thresholds. Secondly, compared to pairwise-like adaptivity, higher-order adaptivity counter-intuitively induces completely opposing emergent phenomena. The former enhances or induces the bistable region, while the latter reduces or even eliminates it. Moreover, although pairwise-like adaptivity triggers discontinuous transitions, higher-order adaptivity causes that phase transitions change from discontinuous to continuous. These findings, both theoretical and experimental, highlight the unique characteristics of higher-order adaptivity in complex systems. Figure 1: Higher-order adaptive spreading model. It describes the co-evolution of higher-order spreading and network topology. Specifically, the model incorporates higher-order adaptivity, characterized by power-law breaking rate of hyperedges and selected fitness, to depict the evolution of higher-order network induced by higher-order spreading dynamics. In turn, network evolution alters the topological structure, which influences the outcomes of spreading dynamics occurring on top of it. Higher-Order Adaptive Spreading Model. Our model is composed of two mutually interacting processes including higher-order spreading and topological evolution, as shown in Fig. 1 . Firstly, we consider a typical epidemic spreading process on hypergraphs, where hyperedges of size m encode higher-order interactions among m individuals. Each node has two states: infected ( I ) and susceptible ( S ). Infected individuals recover at rate \mu , while transmitting the disease to susceptible ones at a certain infection rate. Noteworthy, recent works demonstrate the ubiquity of higher-order spreading mechanisms, which manifest as the nonlinear relationships between infection rate and the exposure to infected contacts St-Onge et al. 2021 ; St-Onge et al. 2022 ; Ferraz de Arruda et al. 2024 . Here, we adopt the power-law infection rate \theta(j;\beta,v)=\beta j^{v} (1) where j is the number of infected nodes in the hyperedge, \beta is basic infection rate and v adjusts nonlinearity. Next, we introduce higher-order adaptivity to characterize how hypergraph evolves with the dynamical states of nodes. In real-world scenarios of disease transmission, people tend to adopt risk-avoidance behaviors, such as avoiding contacts with infected individuals and avoiding participating in groups with infection risks Gross et al. 2006 ; Pastor-Satorras et al. 2015 . Institutions, such as school and workplace, would take lockdown measures to avoid high-risk group interactions, as was the case during the Covid-19 pandemic Lai et al. 2020 ; Perra 2021 ; Flaxman et al. 2020 . These behaviors and measures can be modeled as the decomposition of groups containing infected members, i.e., hyperedge breaking. Empirical observations suggest that the likelihood of taking actions may nonlinearly rise as the number of infections increases. Therefore, we assume that a hyperedge containing j infected nodes and at least 1 susceptible node breaks with a power-law rate \pi(j;r,h)=rj^{h} (2) where r is the

Read the complete text