Emergence Transformer: Dynamical Temporal Attention Matters

Bo-Wei Qin; Kai Du; Wei Lin; Zihan Zhou

arxiv: 2604.19816 · v1 · submitted 2026-04-18 · 💻 cs.AI

Emergence Transformer: Dynamical Temporal Attention Matters

Zihan Zhou , Bo-Wei Qin , Kai Du , Wei Lin This is my paper

Pith reviewed 2026-05-10 07:48 UTC · model grok-4.3

classification 💻 cs.AI

keywords emergence transformerdynamical temporal attentionoscillatory coherenceHopfield networkcontinual learningsocial coherencetemporal sequencesnetwork dynamics

0 comments

The pith

An Emergence Transformer with time-varying attention matrices controls promotion or suppression of emergent coherence in networks.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes an Emergence Transformer built on dynamical temporal attention, where query, key, and value matrices vary with time to create kernels that let network components interact with their own or neighbors' past states. This setup is shown to enable deliberate promotion or suppression of collective coherence, such as oscillatory patterns in complex systems. The authors report that neighbor-focused attention reliably boosts coherence while self-focused attention peaks at an optimal strength due to non-monotonic network effects. Demonstrations include modulating agreement versus plurality in social groups and producing continual learning without forgetting in a Hopfield network. If the mechanism holds, attention alone becomes a tunable handle on emergence across temporal sequences without added constraints.

Core claim

By designing dynamical temporal attention (DTA) with time-varying query, key, and value matrices, we propose an Emergence Transformer. This architecture allows each component to interact with its own or its neighbors' past states through dynamical attention kernels, thereby enabling the promotion and/or suppression of the emergent coherence of components. Neighbor-DTA consistently promotes oscillatory coherence, whereas self-DTA exhibits an optimal attention weight for coherence enhancement, owing to its non-monotonic dependence on network structure. Practically, we demonstrate how DTA reshapes social coherence, suggesting strategies to either enhance agreement or preserve plurality. We also

What carries the argument

Dynamical Temporal Attention (DTA) formed by time-varying query, key, and value matrices that produce attention kernels mediating interactions with past states.

If this is right

Neighbor-DTA promotes oscillatory coherence across the network.
Self-DTA shows non-monotonic dependence on structure, with a peak weight that maximizes coherence enhancement.
DTA can reshape social networks to increase agreement or maintain plurality of views.
DTA applied to Hopfield networks produces emergent continual learning without catastrophic forgetting.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same DTA construction could be tested on climate or biophysical oscillator models to check whether coherence can be tuned solely through attention weights.
The non-monotonic self-attention result implies that topology and attention interact in ways that might allow targeted interventions in other networked dynamical systems.
Findings on social coherence point to possible use of DTA-style modules for designing online platforms that balance consensus and diversity.

Load-bearing premise

Time-varying attention kernels created from dynamical Q, K, V matrices will produce controllable promotion or suppression of coherence without extra fitting, constraints, or domain tuning.

What would settle it

Running the same oscillatory network simulation with neighbor-DTA versus self-DTA and finding that coherence levels do not increase under neighbor attention or lack an optimal peak under self-attention.

read the original abstract

The Transformer, a breakthrough architecture in artificial intelligence, owes its success to the attention mechanism, which utilizes long-range interactions in sequential data, enabling the emergent coherence between large language models (LLMs) and data distributions. However, temporal attention, that is, different forms of long-range interactions in temporal sequences, has rarely been explored in emergence phenomenon of complex systems including oscillatory coherence in quantum, biophysical, or climate systems. Here, by designing dynamical temporal attention (DTA) with time-varying query, key, and value matrices, we propose an Emergence Transformer. This architecture allows each component to interact with its own or its neighbors' past states through dynamical attention kernels, thereby enabling the promotion and/or suppression of the emergent coherence of components. Interestingly, we uncover that neighbor-DTA consistently promotes oscillatory coherence, whereas self-DTA exhibits an optimal attention weight for coherence enhancement, owing to its non-monotonic dependence on network structure. Practically, we demonstrate how DTA reshapes social coherence, suggesting strategies to either enhance agreement or preserve plurality. We further apply DTA to the paradigmatic Hopfield neural network, achieving emergent continual learning without catastrophic forgetting. Together, these results lay a foundation and provide an immediate paradigm for modulating emergence phenomenon in networked dynamics only using DTA.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper frames time-varying attention kernels as a way to tune emergent coherence in networks, with neighbor-DTA promoting it and self-DTA showing a network-dependent optimum, plus a Hopfield demo for continual learning.

read the letter

The core move is to make query, key, and value matrices evolve with time so that nodes in a network can attend to their own or neighbors' histories through these dynamical kernels. This is presented as a general lever for raising or lowering collective oscillatory behavior. They report that neighbor attention reliably increases coherence while self-attention peaks at an intermediate weight because of non-monotonic dependence on topology. The Hopfield example is the most concrete part: they claim the same mechanism produces continual learning without catastrophic forgetting. That result, if reproducible, is the clearest practical payoff. The social-coherence application is sketched more lightly as a way to either drive consensus or maintain diversity. What the work does cleanly is import the transformer primitive into a dynamical-systems setting and show that the self-versus-neighbor distinction matters for the sign of the effect. The non-monotonic self-DTA finding is at least stated as an empirical observation rather than an assumption. The main limitation is that the abstract supplies no explicit update equations, no parameter counts, and no error bars or baseline comparisons, so it is still unclear how much of the reported behavior is generic versus tied to the particular network models they chose. Without those details it is hard to judge whether the kernels are truly controllable without extra fitting. The paper is aimed at people who already work on emergence in physics, biology, or multi-agent systems and who are comfortable treating attention as a modeling tool rather than a pure NLP device. It is not yet a finished architecture paper, but the idea is distinct enough from standard temporal attention variants that a referee could usefully check the derivations and the Hopfield experiments. I would send it out for review.

Referee Report

2 major / 2 minor

Summary. The paper introduces the Emergence Transformer, which incorporates dynamical temporal attention (DTA) defined via time-varying query, key, and value matrices. This architecture enables components in networked systems to interact with their own or neighbors' past states through dynamical attention kernels, with the goal of promoting or suppressing emergent coherence. The central claims are that neighbor-DTA consistently promotes oscillatory coherence while self-DTA exhibits an optimal attention weight due to non-monotonic dependence on network structure; applications include reshaping social coherence and achieving continual learning in Hopfield networks without catastrophic forgetting.

Significance. If the results hold with rigorous support, the work offers a potentially useful bridge between transformer attention mechanisms and the control of emergence in complex networked dynamical systems, with possible applications in social dynamics and neural network continual learning. The explicit attempt to derive controllable coherence modulation from time-varying kernels is a conceptual strength, though its generality remains to be established.

major comments (2)

Abstract: the assertion that neighbor-DTA 'consistently promotes' oscillatory coherence and that self-DTA has an 'optimal' weight is presented without any defining equations for the time-varying Q/K/V matrices, without any reported data or error bars, and without derivation steps; these claims are load-bearing for the central assertion that DTA enables controllable promotion/suppression of emergent coherence.
Abstract: the reported non-monotonic dependence of the self-DTA optimum on network structure is stated as an empirical finding, but no specific network model, Hamiltonian, or simulation protocol is supplied to show whether the optimum arises from first principles or from fitting the same data used to demonstrate the effect.

minor comments (2)

The abstract introduces 'dynamical attention kernels' and 'time-varying query, key, and value matrices' without a compact mathematical definition or reference to the section where the update rules are given; this notation should be clarified early in the methods.
No mention is made of baseline comparisons (standard attention, static kernels, or mean-field approximations) that would be needed to isolate the contribution of the dynamical component.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their careful reading and valuable comments, which have prompted us to clarify several aspects of our work. We respond to each major comment below and indicate the revisions made to the manuscript.

read point-by-point responses

Referee: Abstract: the assertion that neighbor-DTA 'consistently promotes' oscillatory coherence and that self-DTA has an 'optimal' weight is presented without any defining equations for the time-varying Q/K/V matrices, without any reported data or error bars, and without derivation steps; these claims are load-bearing for the central assertion that DTA enables controllable promotion/suppression of emergent coherence.

Authors: The abstract provides a high-level overview of our results, while the detailed definitions of the time-varying Q, K, and V matrices, along with the derivation of the dynamical temporal attention mechanism, are presented in Section 2. The supporting simulation data, including error bars from repeated trials, are shown in Figures 4 and 5. We acknowledge that the abstract could better signpost these elements. In the revised manuscript, we have updated the abstract to include a reference to the methods and results sections where the equations, derivations, and data are provided. This maintains the abstract's conciseness while addressing the concern about the load-bearing claims. revision: yes
Referee: Abstract: the reported non-monotonic dependence of the self-DTA optimum on network structure is stated as an empirical finding, but no specific network model, Hamiltonian, or simulation protocol is supplied to show whether the optimum arises from first principles or from fitting the same data used to demonstrate the effect.

Authors: We agree with the referee that explicit details on the underlying models are important for interpreting the empirical finding. The non-monotonic dependence is observed in simulations of specific networked systems, and we have now included in the revised manuscript a clearer description of the network models (coupled phase oscillators for the social coherence application and the standard Hopfield energy function for the continual learning case), the simulation protocols, and how the attention weight is varied independently of the data used for demonstration. This shows that the optimum emerges from the dynamical equations rather than being a result of data fitting. We have also added a brief discussion on the connection to first-principles modeling of attention in dynamical systems. revision: yes

Circularity Check

0 steps flagged

No significant circularity; proposal and empirical demonstration are self-contained

full rationale

The paper defines the Emergence Transformer by introducing dynamical temporal attention (DTA) via time-varying Q/K/V matrices, then applies it to social networks and Hopfield models to observe coherence modulation. These are presented as design choices and simulation results rather than derivations that reduce to prior fits or self-citations. The non-monotonic optimal weight for self-DTA is described as an observed dependence on network structure, not a fitted parameter renamed as prediction. No load-bearing step equates outputs to inputs by construction, and the central claims rest on the explicit architecture definition plus external model applications.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 2 invented entities

The abstract relies on the unproven premise that dynamical kernels built from time-varying Q/K/V matrices can directly control emergent coherence; no free parameters are explicitly named, but the existence of an 'optimal' weight implies at least one tunable scalar per network.

free parameters (1)

self-DTA attention weight
Described as having an optimal value that depends on network structure; the optimum is presented as a discovered feature rather than a derived constant.

axioms (1)

domain assumption Time-varying query, key, and value matrices can be defined such that their kernels interact with past states of self or neighbors.
This is the foundational design choice stated in the abstract without further justification.

invented entities (2)

Emergence Transformer no independent evidence
purpose: Architecture that modulates emergent coherence via DTA
New named model introduced in the abstract; no independent evidence supplied.
dynamical temporal attention (DTA) kernels no independent evidence
purpose: Time-dependent interaction mechanism for coherence control
Core new mechanism postulated to produce the reported promotion/suppression effects.

pith-pipeline@v0.9.0 · 5522 in / 1654 out tokens · 48816 ms · 2026-05-10T07:48:40.496097+00:00 · methodology

discussion (0)

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Attention by Synchronization in Coupled Oscillator Networks
cs.LG 2026-06 unverdicted novelty 7.0

Kuramoto synchronization dynamics implement a provably unique and globally attractive attention mechanism that replaces softmax for physical substrates and shows competitive empirical performance.

Reference graph

Works this paper leans on

68 extracted references · 68 canonical work pages · cited by 1 Pith paper

[1]

Vaswani, N

A. Vaswani, N. Shazeer, N. Parmar, et al. Attention is all you need. InProceedings of the 31st International Con- ference on Neural Information Processing Systems, pages 6000–6010, 2017

work page 2017
[2]

H. R. Kirk, B. Vidgen, P. R¨ ottger, et al. The benefits, risks and bounds of personalizing the alignment of large language models to individuals.Nature Machine Intelligence, 6:383– 392, 2024

work page 2024
[3]

Pikovsky, M

A. Pikovsky, M. Rosenblum, and J. Kurths.Synchroniza- tion: A Universal Concept in Nonlinear Sciences. Cam- bridge University Press, 2010

work page 2010
[4]

M. L. Wong, C. E. Cleland, D. Jr. Arena, et al. On the roles of function and selection in evolving systems.Proceed- ings of the National Academy Sciences of the United States of America, 120:e2310223120, 2023

work page 2023
[5]

Bloch, A

J. Bloch, A. Cavalleri, V. Galitski, et al. Strongly corre- lated electron–photon systems.Nature, 606:41–48, 2022

work page 2022
[6]

Moille, J

G. Moille, J. Stone, M. Chojnacky, et al. Kerr-induced synchronization of a cavity soliton to an optical reference. Nature, 624:267–274, 2023

work page 2023
[7]

De Domenico

M. De Domenico. More is different in real-world multilayer networks.Nature Physics, 19:1247–1262, 2023

work page 2023
[8]

M. Yan, C. Huang, P. Bienstman, et al. Emerging oppor- tunities and challenges for the future of reservoir comput- ing.Nature Communications, 15:2056, 2024

work page 2056
[9]

Artime, M

O. Artime, M. Grassia, M. De Domenico, et al. Robustness and resilience of complex networks.Nature Review Physics, 6:114–131, 2024

work page 2024
[10]

Chang, D

C.-Y. Chang, D. Baji´ c, J. C. C. Vila, et al. Emergent coexistence in multispecies microbial communities.Science, 381:343–348, 2023

work page 2023
[11]

Raccuglia, R

D. Raccuglia, R. Su´ arez-Grimalt, L. Krumm, et al. Net- work synchrony creates neural filters promoting quiescence inDrosophila.Nature, 646:667–675, 2025

work page 2025
[12]

K. Zhou, Z. Liu, Y. Qiao, et al. Domain generalization: A survey.IEEE Transactions on Pattern Analysis and Ma- chine Intelligence, 45:4396–4415, 2023

work page 2023
[13]

G. V. Osipov, J. Kurths, and C. Zhou.Synchronization in Oscillatory Networks. Springer-Verlag Berlin Heidelberg, 2007

work page 2007
[14]

Sugitani, Y

Y. Sugitani, Y. Zhang, and A. E. Motter. Synchro- nizing chaos with imperfections.Physical Review Letters, 126:164101, 2021

work page 2021
[15]

K´ om´ ar, E

P. K´ om´ ar, E. Kessler, M. Bishof, et al. A quantum net- work of clocks.Nature Physics, 10:582–587, 2014

work page 2014
[16]

Rohden, A

M. Rohden, A. Sorge, M. Timme, et al. Self-organized synchronization in decentralized power grids.Physical Re- view Letters, 109:064101, 2012

work page 2012
[17]

Smith, O

O. Smith, O. Cattell, E. Farcot, et al. The effect of re- newable energy incorporation on power grid stability and resilience.Science Advances, 8:eabj6734, 2022

work page 2022
[18]

F. Min, C. Chen, and N. G. R. Broderick. Coupled homo- geneous hopfield neural networks: Simplest model design, synchronization, and multiplierless circuit implementation. IEEE Transactions on Neural Networks and Learning Sys- tems, 36:11632–11639, 2025

work page 2025
[19]

W. Fu, Z. Li, W. Lin, et al. The role of higher-order self- dynamics in neural dynamical networks: Preserving mem- ory capacity and enhancing retrieval basin.SIAM Journal on Applied Mathematics, 85:1834–1855, 2025

work page 2025
[20]

X. Liu, D. Zhang, and X. He. Unveiling the role of climate 11 in spatially synchronized locust outbreak risks.Science Ad- vances, 10:eadj1164, 2024

work page 2024
[21]

X. Ren, A. Brodovskaya, and J. L. Hudson. Connec- tivity and neuronal synchrony during seizures.Journal of Neuroscience, 41:7623–7635, 2021

work page 2021
[22]

di Santo, P

S. di Santo, P. Villegas, R. Burioni, et al. Landau– ginzburg theory of cortex dynamics: Scale-free avalanches emerge at the edge of synchronization.Proceedings of the National Academy of Sciences of the United States of Amer- ica, 115:E1356–E1365, 2018

work page 2018
[23]

Kuramoto

Y. Kuramoto. Self-entrainment of a population of cou- pled non-linear oscillators. In H. Araki, editor,Interna- tional Symposium on Mathematical Problems in Theoretical Physics, pages 420–422. Springer, Berlin, Heidelberg, 1975

work page 1975
[24]

Zhang, P

Y. Zhang, P. S. Skardal, F. Battiston, et al. Deeper but smaller: Higher-order interactions increase linear stability but shrink basins.Science Advances, 10:eado8049, 2024

work page 2024
[25]

A. P. Mill´ an, H. Sun, L. Giambagli, et al. Topology shapes dynamics of higher-order networks.Nature Physics, 21:353–361, 2025

work page 2025
[26]

Zhang and S

Y. Zhang and S. H. Strogatz. Designing temporal net- works that synchronize under resource constraints.Nature Communications, 12:3273, 2021

work page 2021
[27]

Nijholt, J

E. Nijholt, J. L. Ocampo-Espindola, D. Eroglu, et al. Emergent hypernetworks in weakly coupled oscillators.Na- ture Communications, 13:4849, 2022

work page 2022
[28]

Zhong, W

Z. Zhong, W. Lin, and B.-W. Qin. Modulating biological rhythms: A noncomputational strategy harnessing nonlin- earity and decoupling frequency and amplitude.Physical Review Letters, 131:138401, 2023

work page 2023
[29]

J. A. Acebr´ on, L. L. Bonilla, C. J. P´ erez Vicente, et al. The kuramoto model: A simple paradigm for synchroniza- tion phenomena.Review of Modern Physics, 77:137, 2005

work page 2005
[30]

S. H. Strogatz. From kuramoto to crawford: exploring the onset of synchronization in populations of coupled os- cillators.Physica D, 143:1–20, 2000

work page 2000
[31]

G. S. Medvedev. Small-world networks of kuramoto os- cillators.Physica D, 266:13–22, 2014

work page 2014
[32]

M. A. Gkogkas and C. Kuehn. Graphop mean-field lim- its for kuramoto-type models.SIAM Journal on Applied Dynamical Systems, 21:248–283, 2022

work page 2022
[33]

Ott and T

E. Ott and T. M. Antonson. Low dimensional behavior of large systems of globally coupled oscillators.Chaos: An Interdisciplinary Journal of Nonlinear Science, 18:037113, 2008

work page 2008
[34]

O. E. Omel’chenko. Periodic orbits in the ott-antonsen manifold.Nonlinearity, 36:845, 2022

work page 2022
[35]

Motter, S

A. Motter, S. Myers, M. Anghel, et al. Spontaneous synchrony in power-grid networks.Nature Physics, 9:191– 197, 2013

work page 2013
[36]

Nazerian, J

A. Nazerian, J. D. Hart, M. Lodi, et al. The efficiency of synchronization dynamics and the role of network syncre- activity.Nature Communications, 15:9003, 2024

work page 2024
[37]

S. Lee, L. J. Kuklinski, and M. Timme. Extreme syn- chronization transitions.Nature Communications, 16:4505, 2025

work page 2025
[38]

Buend´ ıa

V. Buend´ ıa. Mesoscopic theory for coupled stochastic oscillators.Physical Review Letters, 134:197201, 2025

work page 2025
[39]

Appeltant, M

L. Appeltant, M. Soriano, G. Van der Sande, et al. Infor- mation processing using a single dynamical node as complex system.Nature Communications, 2:468, 2011

work page 2011
[40]

X.-Y. Duan, X. Ying, S. Leng, et al. Embedding theory of reservoir computing and reducing reservoir network using time delays.Physical Review Research, 5:L022041, 2023

work page 2023
[41]

Bena¨ ım, M

M. Bena¨ ım, M. Ledoux, and O. Raimond. Self- interacting diffusions.Probability Theory and Related Fields, 122:1–41, 2002

work page 2002
[42]

Candia, C

C. Candia, C. Jara-Figueroa, C. Rodriguez-Sickert, et al. The universal decay of collective memory and attention. Nature Human Behaviour, 3:82–91, 2019

work page 2019
[43]

Watts and S

D. Watts and S. Strogatz. Collective dynamics of ‘small- world’ networks.Nature, 393:440–442, 1998

work page 1998
[44]

Strogatz

S. Strogatz. Exploring complex networks.Nature, 410:268–276, 2001

work page 2001
[45]

Sonnenschein and L

B. Sonnenschein and L. Schimansky-Geier. Onset of syn- chronization in complex networks of noisy oscillators.Phys- ical Review E, 85:051116, 2012

work page 2012
[46]

W. Zou, S. He, D. V. Senthilkumar, et al. Solvable dy- namics of coupled high-dimensional generalized limit-cycle oscillators.Physical Review Letters, 130:107202, 2023

work page 2023
[47]

T. D. Frank, P. J. Beek, and R. Friedrich. Fokker-planck perspective on stochastic delay systems: Exact solutions and data analysis of biological systems.Physical Review E, 68:021912, 2003

work page 2003
[48]

A. Ross, S. N. Kyrychko, K. B. Blyuss, et al. Dynam- ics of coupled kuramoto oscillators with distributed delays. Chaos: An Interdisciplinary Journal of Nonlinear Science, 31:103107, 2021

work page 2021
[49]

Erd˝ os and A

P. Erd˝ os and A. R´ enyi. On the evolution of random graphs.Publication of the Mathematical Institute of the Hungarian Academy of Sciences, 5:17–60, 1960

work page 1960
[50]

Barab´ asi and R

A.-L. Barab´ asi and R. Albert. Emergence of scaling in random networks.Science, 286:509–512, 1999

work page 1999
[51]

J. Ojer, M. Starnini, and R. Pastor-Satorras. Modeling explosive opinion depolarization in interdependent topics. Physical Review Letters, 130:207401, 2023

work page 2023
[52]

Rossi and N

R. Rossi and N. Ahmed. The network data repository with interactive graph analytics and visualization. InPro- ceedings of the AAAI Conference on Artificial Intelligence, volume 29, 2015

work page 2015
[53]

T. Lin, Y. Wang, X. Liu, and X. Qiu. A survey of trans- formers.AI Open, 3:111–132, 2022

work page 2022
[54]

T. D. Frank. Delay fokker-planck equations, perturbation theory, and data analysis for nonlinear stochastic systems with time delays.Physical Review E, 71:031106, 2005

work page 2005
[55]

W. S. Lee, E. Ott, and T. M. Antonsen. Large cou- pled oscillator systems with heterogeneous interaction de- lays.Physical Review Letters, 103:044101, 2009

work page 2009
[56]

Zhang, J

Y. Zhang, J. L. Ocampo-Espindola, I. Z. Kiss, et al. Ran- dom heterogeneity outperforms design in network synchro- nization.Proceedings of the National Academy of Sciences of the United States of America, 118:e2024299118, 2021

work page 2021
[57]

C. Cai, J. Yu, X. Zhang, et al. A model for propagation of rna structural memory through biomolecular condensates. Nature Cell Biology, 27:1381–1386, 2025

work page 2025
[58]

Schmolke and E

F. Schmolke and E. Lutz. Noise-induced quantum syn- chronization.Physical Review Letters, 129:250601, 2022

work page 2022
[59]

de Vega and D

I. de Vega and D. Alonso. Dynamics of non-markovian open quantum systems.Review of Modern Physics, 89:015001, 2017

work page 2017
[60]

F. Takens. Detecting strange attractors in turbulence. Lecture Notes in Mathematics, 898:366–381, 1981

work page 1981
[61]

H. Ma, S. Leng, K. Aihara, et al. Randomly distributed embedding making short-term high-dimensional data pre- dictable.Proceedings of the National Academy of Sciences of the United States of America, 115:E9994–E10002, 2018

work page 2018
[62]

R. V. Raut, Z. P. Rosenthal, X. Wang, et al. Arousal as a universal embedding for spatiotemporal brain dynamics. 12 Nature, 2025

work page 2025
[63]

J. Liu, J. Zhang, and Y. Wang. Secure communication via chaotic synchronization based on reservoir computing. IEEE Transactions on Neural Networks and Learning Sys- tems, 35:285–299, 2024

work page 2024
[64]

N. E. Friedkin and E. C. Johnsen. Social influence and opinions.The Journal of Mathematical Sociology, 15:193– 206, 1990

work page 1990
[65]

Nishikawa, Y.-C

T. Nishikawa, Y.-C. Lai, and F. C. Hoppensteadt. Capac- ity of oscillatory associative-memory networks with error- free retrieval.Physical Review Letters, 92:108101, 2004

work page 2004
[66]

S. P. Cornelius, W. L. Kath, and A. E. Motter. Realis- tic control of network dynamics.Nature Communications, 4:1942, 2013

work page 1942
[67]

F. C. Hoppensteadt and E. M. Izhikevich.Weakly Con- nected Neural Networks. Springer, 1997

work page 1997
[68]

Botet, R

R. Botet, R. Jullien, and P. Pfeuty. Size scaling for infinitely coordinated systems.Physical Review Letters, 49:478, 1982. 13 METHODS DTA in a discrete-time model In order to have a better understanding and compari- son with the classical Transformer architecture, we also demonstrate how to derive attention information for up- dating the phase states whe...

work page 1982

[1] [1]

Vaswani, N

A. Vaswani, N. Shazeer, N. Parmar, et al. Attention is all you need. InProceedings of the 31st International Con- ference on Neural Information Processing Systems, pages 6000–6010, 2017

work page 2017

[2] [2]

H. R. Kirk, B. Vidgen, P. R¨ ottger, et al. The benefits, risks and bounds of personalizing the alignment of large language models to individuals.Nature Machine Intelligence, 6:383– 392, 2024

work page 2024

[3] [3]

Pikovsky, M

A. Pikovsky, M. Rosenblum, and J. Kurths.Synchroniza- tion: A Universal Concept in Nonlinear Sciences. Cam- bridge University Press, 2010

work page 2010

[4] [4]

M. L. Wong, C. E. Cleland, D. Jr. Arena, et al. On the roles of function and selection in evolving systems.Proceed- ings of the National Academy Sciences of the United States of America, 120:e2310223120, 2023

work page 2023

[5] [5]

Bloch, A

J. Bloch, A. Cavalleri, V. Galitski, et al. Strongly corre- lated electron–photon systems.Nature, 606:41–48, 2022

work page 2022

[6] [6]

Moille, J

G. Moille, J. Stone, M. Chojnacky, et al. Kerr-induced synchronization of a cavity soliton to an optical reference. Nature, 624:267–274, 2023

work page 2023

[7] [7]

De Domenico

M. De Domenico. More is different in real-world multilayer networks.Nature Physics, 19:1247–1262, 2023

work page 2023

[8] [8]

M. Yan, C. Huang, P. Bienstman, et al. Emerging oppor- tunities and challenges for the future of reservoir comput- ing.Nature Communications, 15:2056, 2024

work page 2056

[9] [9]

Artime, M

O. Artime, M. Grassia, M. De Domenico, et al. Robustness and resilience of complex networks.Nature Review Physics, 6:114–131, 2024

work page 2024

[10] [10]

Chang, D

C.-Y. Chang, D. Baji´ c, J. C. C. Vila, et al. Emergent coexistence in multispecies microbial communities.Science, 381:343–348, 2023

work page 2023

[11] [11]

Raccuglia, R

D. Raccuglia, R. Su´ arez-Grimalt, L. Krumm, et al. Net- work synchrony creates neural filters promoting quiescence inDrosophila.Nature, 646:667–675, 2025

work page 2025

[12] [12]

K. Zhou, Z. Liu, Y. Qiao, et al. Domain generalization: A survey.IEEE Transactions on Pattern Analysis and Ma- chine Intelligence, 45:4396–4415, 2023

work page 2023

[13] [13]

G. V. Osipov, J. Kurths, and C. Zhou.Synchronization in Oscillatory Networks. Springer-Verlag Berlin Heidelberg, 2007

work page 2007

[14] [14]

Sugitani, Y

Y. Sugitani, Y. Zhang, and A. E. Motter. Synchro- nizing chaos with imperfections.Physical Review Letters, 126:164101, 2021

work page 2021

[15] [15]

K´ om´ ar, E

P. K´ om´ ar, E. Kessler, M. Bishof, et al. A quantum net- work of clocks.Nature Physics, 10:582–587, 2014

work page 2014

[16] [16]

Rohden, A

M. Rohden, A. Sorge, M. Timme, et al. Self-organized synchronization in decentralized power grids.Physical Re- view Letters, 109:064101, 2012

work page 2012

[17] [17]

Smith, O

O. Smith, O. Cattell, E. Farcot, et al. The effect of re- newable energy incorporation on power grid stability and resilience.Science Advances, 8:eabj6734, 2022

work page 2022

[18] [18]

F. Min, C. Chen, and N. G. R. Broderick. Coupled homo- geneous hopfield neural networks: Simplest model design, synchronization, and multiplierless circuit implementation. IEEE Transactions on Neural Networks and Learning Sys- tems, 36:11632–11639, 2025

work page 2025

[19] [19]

W. Fu, Z. Li, W. Lin, et al. The role of higher-order self- dynamics in neural dynamical networks: Preserving mem- ory capacity and enhancing retrieval basin.SIAM Journal on Applied Mathematics, 85:1834–1855, 2025

work page 2025

[20] [20]

X. Liu, D. Zhang, and X. He. Unveiling the role of climate 11 in spatially synchronized locust outbreak risks.Science Ad- vances, 10:eadj1164, 2024

work page 2024

[21] [21]

X. Ren, A. Brodovskaya, and J. L. Hudson. Connec- tivity and neuronal synchrony during seizures.Journal of Neuroscience, 41:7623–7635, 2021

work page 2021

[22] [22]

di Santo, P

S. di Santo, P. Villegas, R. Burioni, et al. Landau– ginzburg theory of cortex dynamics: Scale-free avalanches emerge at the edge of synchronization.Proceedings of the National Academy of Sciences of the United States of Amer- ica, 115:E1356–E1365, 2018

work page 2018

[23] [23]

Kuramoto

Y. Kuramoto. Self-entrainment of a population of cou- pled non-linear oscillators. In H. Araki, editor,Interna- tional Symposium on Mathematical Problems in Theoretical Physics, pages 420–422. Springer, Berlin, Heidelberg, 1975

work page 1975

[24] [24]

Zhang, P

Y. Zhang, P. S. Skardal, F. Battiston, et al. Deeper but smaller: Higher-order interactions increase linear stability but shrink basins.Science Advances, 10:eado8049, 2024

work page 2024

[25] [25]

A. P. Mill´ an, H. Sun, L. Giambagli, et al. Topology shapes dynamics of higher-order networks.Nature Physics, 21:353–361, 2025

work page 2025

[26] [26]

Zhang and S

Y. Zhang and S. H. Strogatz. Designing temporal net- works that synchronize under resource constraints.Nature Communications, 12:3273, 2021

work page 2021

[27] [27]

Nijholt, J

E. Nijholt, J. L. Ocampo-Espindola, D. Eroglu, et al. Emergent hypernetworks in weakly coupled oscillators.Na- ture Communications, 13:4849, 2022

work page 2022

[28] [28]

Zhong, W

Z. Zhong, W. Lin, and B.-W. Qin. Modulating biological rhythms: A noncomputational strategy harnessing nonlin- earity and decoupling frequency and amplitude.Physical Review Letters, 131:138401, 2023

work page 2023

[29] [29]

J. A. Acebr´ on, L. L. Bonilla, C. J. P´ erez Vicente, et al. The kuramoto model: A simple paradigm for synchroniza- tion phenomena.Review of Modern Physics, 77:137, 2005

work page 2005

[30] [30]

S. H. Strogatz. From kuramoto to crawford: exploring the onset of synchronization in populations of coupled os- cillators.Physica D, 143:1–20, 2000

work page 2000

[31] [31]

G. S. Medvedev. Small-world networks of kuramoto os- cillators.Physica D, 266:13–22, 2014

work page 2014

[32] [32]

M. A. Gkogkas and C. Kuehn. Graphop mean-field lim- its for kuramoto-type models.SIAM Journal on Applied Dynamical Systems, 21:248–283, 2022

work page 2022

[33] [33]

Ott and T

E. Ott and T. M. Antonson. Low dimensional behavior of large systems of globally coupled oscillators.Chaos: An Interdisciplinary Journal of Nonlinear Science, 18:037113, 2008

work page 2008

[34] [34]

O. E. Omel’chenko. Periodic orbits in the ott-antonsen manifold.Nonlinearity, 36:845, 2022

work page 2022

[35] [35]

Motter, S

A. Motter, S. Myers, M. Anghel, et al. Spontaneous synchrony in power-grid networks.Nature Physics, 9:191– 197, 2013

work page 2013

[36] [36]

Nazerian, J

A. Nazerian, J. D. Hart, M. Lodi, et al. The efficiency of synchronization dynamics and the role of network syncre- activity.Nature Communications, 15:9003, 2024

work page 2024

[37] [37]

S. Lee, L. J. Kuklinski, and M. Timme. Extreme syn- chronization transitions.Nature Communications, 16:4505, 2025

work page 2025

[38] [38]

Buend´ ıa

V. Buend´ ıa. Mesoscopic theory for coupled stochastic oscillators.Physical Review Letters, 134:197201, 2025

work page 2025

[39] [39]

Appeltant, M

L. Appeltant, M. Soriano, G. Van der Sande, et al. Infor- mation processing using a single dynamical node as complex system.Nature Communications, 2:468, 2011

work page 2011

[40] [40]

X.-Y. Duan, X. Ying, S. Leng, et al. Embedding theory of reservoir computing and reducing reservoir network using time delays.Physical Review Research, 5:L022041, 2023

work page 2023

[41] [41]

Bena¨ ım, M

M. Bena¨ ım, M. Ledoux, and O. Raimond. Self- interacting diffusions.Probability Theory and Related Fields, 122:1–41, 2002

work page 2002

[42] [42]

Candia, C

C. Candia, C. Jara-Figueroa, C. Rodriguez-Sickert, et al. The universal decay of collective memory and attention. Nature Human Behaviour, 3:82–91, 2019

work page 2019

[43] [43]

Watts and S

D. Watts and S. Strogatz. Collective dynamics of ‘small- world’ networks.Nature, 393:440–442, 1998

work page 1998

[44] [44]

Strogatz

S. Strogatz. Exploring complex networks.Nature, 410:268–276, 2001

work page 2001

[45] [45]

Sonnenschein and L

B. Sonnenschein and L. Schimansky-Geier. Onset of syn- chronization in complex networks of noisy oscillators.Phys- ical Review E, 85:051116, 2012

work page 2012

[46] [46]

W. Zou, S. He, D. V. Senthilkumar, et al. Solvable dy- namics of coupled high-dimensional generalized limit-cycle oscillators.Physical Review Letters, 130:107202, 2023

work page 2023

[47] [47]

T. D. Frank, P. J. Beek, and R. Friedrich. Fokker-planck perspective on stochastic delay systems: Exact solutions and data analysis of biological systems.Physical Review E, 68:021912, 2003

work page 2003

[48] [48]

A. Ross, S. N. Kyrychko, K. B. Blyuss, et al. Dynam- ics of coupled kuramoto oscillators with distributed delays. Chaos: An Interdisciplinary Journal of Nonlinear Science, 31:103107, 2021

work page 2021

[49] [49]

Erd˝ os and A

P. Erd˝ os and A. R´ enyi. On the evolution of random graphs.Publication of the Mathematical Institute of the Hungarian Academy of Sciences, 5:17–60, 1960

work page 1960

[50] [50]

Barab´ asi and R

A.-L. Barab´ asi and R. Albert. Emergence of scaling in random networks.Science, 286:509–512, 1999

work page 1999

[51] [51]

J. Ojer, M. Starnini, and R. Pastor-Satorras. Modeling explosive opinion depolarization in interdependent topics. Physical Review Letters, 130:207401, 2023

work page 2023

[52] [52]

Rossi and N

R. Rossi and N. Ahmed. The network data repository with interactive graph analytics and visualization. InPro- ceedings of the AAAI Conference on Artificial Intelligence, volume 29, 2015

work page 2015

[53] [53]

T. Lin, Y. Wang, X. Liu, and X. Qiu. A survey of trans- formers.AI Open, 3:111–132, 2022

work page 2022

[54] [54]

T. D. Frank. Delay fokker-planck equations, perturbation theory, and data analysis for nonlinear stochastic systems with time delays.Physical Review E, 71:031106, 2005

work page 2005

[55] [55]

W. S. Lee, E. Ott, and T. M. Antonsen. Large cou- pled oscillator systems with heterogeneous interaction de- lays.Physical Review Letters, 103:044101, 2009

work page 2009

[56] [56]

Zhang, J

Y. Zhang, J. L. Ocampo-Espindola, I. Z. Kiss, et al. Ran- dom heterogeneity outperforms design in network synchro- nization.Proceedings of the National Academy of Sciences of the United States of America, 118:e2024299118, 2021

work page 2021

[57] [57]

C. Cai, J. Yu, X. Zhang, et al. A model for propagation of rna structural memory through biomolecular condensates. Nature Cell Biology, 27:1381–1386, 2025

work page 2025

[58] [58]

Schmolke and E

F. Schmolke and E. Lutz. Noise-induced quantum syn- chronization.Physical Review Letters, 129:250601, 2022

work page 2022

[59] [59]

de Vega and D

I. de Vega and D. Alonso. Dynamics of non-markovian open quantum systems.Review of Modern Physics, 89:015001, 2017

work page 2017

[60] [60]

F. Takens. Detecting strange attractors in turbulence. Lecture Notes in Mathematics, 898:366–381, 1981

work page 1981

[61] [61]

H. Ma, S. Leng, K. Aihara, et al. Randomly distributed embedding making short-term high-dimensional data pre- dictable.Proceedings of the National Academy of Sciences of the United States of America, 115:E9994–E10002, 2018

work page 2018

[62] [62]

R. V. Raut, Z. P. Rosenthal, X. Wang, et al. Arousal as a universal embedding for spatiotemporal brain dynamics. 12 Nature, 2025

work page 2025

[63] [63]

J. Liu, J. Zhang, and Y. Wang. Secure communication via chaotic synchronization based on reservoir computing. IEEE Transactions on Neural Networks and Learning Sys- tems, 35:285–299, 2024

work page 2024

[64] [64]

N. E. Friedkin and E. C. Johnsen. Social influence and opinions.The Journal of Mathematical Sociology, 15:193– 206, 1990

work page 1990

[65] [65]

Nishikawa, Y.-C

T. Nishikawa, Y.-C. Lai, and F. C. Hoppensteadt. Capac- ity of oscillatory associative-memory networks with error- free retrieval.Physical Review Letters, 92:108101, 2004

work page 2004

[66] [66]

S. P. Cornelius, W. L. Kath, and A. E. Motter. Realis- tic control of network dynamics.Nature Communications, 4:1942, 2013

work page 1942

[67] [67]

F. C. Hoppensteadt and E. M. Izhikevich.Weakly Con- nected Neural Networks. Springer, 1997

work page 1997

[68] [68]

Botet, R

R. Botet, R. Jullien, and P. Pfeuty. Size scaling for infinitely coordinated systems.Physical Review Letters, 49:478, 1982. 13 METHODS DTA in a discrete-time model In order to have a better understanding and compari- son with the classical Transformer architecture, we also demonstrate how to derive attention information for up- dating the phase states whe...

work page 1982