Kuramoto Attention: Synchronizing Self-Attention on the Torus

arXiv:2606.11585v1 Announce Type: new Abstract: We introduce Kuramoto attention, a self-attention layer in which each hidden coordinate is an angle. The layer scores tokens by gated cosine similarity, attends over previous phase states, and updates each token by the tangent component of the attention-weighted circular mean. Because the values are the raw phase states, this update is exactly the Kuramoto coupling term $\sum_u A_{t,u}\sin(\theta_u-\theta_t)$, with the attention matrix acting as an adaptive, content-dependent coupling kernel. Equivalently, the gated score is a learned metric on the torus that selects which tokens couple, and the update pulls each token toward the circular mean of the tokens it selects, tightening their phase agreement. The same two ingredients, an invariant similarity score and an on-manifold mean, define such a layer on any compact group; the torus is the abelian case, where both are closed-form. The softmax weights solve an entropy-regularized phase-retrieval problem, and rotary position enters as a position-dependent phase drift in the score. On enwiki8 character-level language modeling, the layer trains as a functional language model whose bits-per-character stays close to a strong matched RoPE+SwiGLU transformer: within $0.02$ BPC at one million parameters ($1.637\pm0.010$ versus $1.616\pm0.004$) and level on the median at five million ($1.448$ versus $1.452$ over five seeds) with the transformer ahead on the mean ($1.468$ versus $1.456$). These experiments establish that the constrained geometric structure is a viable language model at this scale; the structure itself, and its synchronization reading, is the contribution. Ablations isolate the load-bearing components, and the result gives a compact bridge between self-attention and phase synchronization.

Abstract:We introduce Kuramoto attention, a self-attention layer in which each hidden coordinate is an angle. The layer scores tokens by gated cosine similarity, attends over previous phase states, and updates each token by the tangent component of the attention-weighted circular mean. Because the values are the raw phase states, this update is exactly the Kuramoto coupling term $\sum_u A_{t,u}\sin(\theta_u-\theta_t)$, with the attention matrix acting as an adaptive, content-dependent coupling kernel. Equivalently, the gated score is a learned metric on the torus that selects which tokens couple, and the update pulls each token toward the circular mean of the tokens it selects, tightening their phase agreement. The same two ingredients, an invariant similarity score and an on-manifold mean, define such a layer on any compact group; the torus is the abelian case, where both are closed-form. The softmax weights solve an entropy-regularized phase-retrieval problem, and rotary position enters as a position-dependent phase drift in the score. On enwiki8 character-level language modeling, the layer trains as a functional language model whose bits-per-character stays close to a strong matched RoPE+SwiGLU transformer: within $0.02$ BPC at one million parameters ($1.637\pm0.010$ versus $1.616\pm0.004$) and level on the median at five million ($1.448$ versus $1.452$ over five seeds) with the transformer ahead on the mean ($1.468$ versus $1.456$). These experiments establish that the constrained geometric structure is a viable language model at this scale; the structure itself, and its synchronization reading, is the contribution. Ablations isolate the load-bearing components, and the result gives a compact bridge between self-attention and phase synchronization.

Machine Learning (cs.LG); Computation and Language (cs.CL); Adaptation and Self-Organizing Systems (nlin.AO)

arXiv:2606.11585 [cs.LG]

(or arXiv:2606.11585v1 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2606.11585 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

analiza aiqbench

Kuramoto Attention: Synchronizing Self-Attention on the Torus

Subiecte conexe