Everything in its right place: attention is a radio head
Organizations: Departament d’Enginyeria Informàtica i Matemàtiques, Universitat Rovira i Virgili, 43007 Tarragona, Spain
Abstract
Attention, the core of language models, is a digital calculation over tokens: an exponential selects them, a division normalizes and a positional code sets the order. We show that attention can instead emerge from the synchronization of a receiver with its context tokens. Each past token is an oscillator whose state carries its content and whose frequency drifts with its age. The present token selects them by content, aligning an antenna with its query, and by position, setting the frequency of a receiver with its dial. Solving Kuramoto's problem for this receiver yields an attention layer in closed form with two phases: above a critical coupling set by the context, the receiver locks and its weights normalize; below it, it stays silent and its weights sum to less than one. Trained networks use both phases to select, abstain and count, fail where the theory predicts and keep reading far beyond the training length. Each part of attention finds its place in one device, a radio head: it tunes in to what to hear and when, and stays silent when nothing relevant is on air.
Figures & tables
| synchronization | attention | radio |
|---|---|---|
| oscillator | token , value (no key) | station |
| frequency | position, age | carrier |
| driven part | reading token | receiver |
| coupling | content match | antenna |
| detuning | position mismatch | dial |
| order parameter | relevance found | signal |
| softmax + RoPE | Pasqualetti–Guo | radio head | |
| state | — | position | content |
| position | rotation of , | anchors | frequency (age) |
| dynamics | none | gradient flow | partial locking |
| weights | Gibbs | fixed-point readout | |
| normalization | locked, silent | ||
| no match | average | average | silence |
| synthetic tasks | language (TinyStories) | |
| layers, heads | 2, 1 | 2, 1 to 8 |
| , feed-forward width | 64, 128 | 64 to 1024, |
| dropout | 0 | 0.1 |
| vocabulary | 51 (recall: 577) | 8004 [ 6 ] |
| training length | 32 (recall: 48) | 128 |
| training data | 20 000 new sequences per epoch | sequences of 128 tokens, one pass |
Attention Mechanisms
Attention mechanisms let neural networks weight information from other tokens or regions according to relevance, enabling context-dependent processing.
See all 541 papers