Kathleen : Oscillator-Based Byte-Level Text Classification Without Tokenization or Attention
George Fountzoulas · 2026
Read Kathleen : Oscillator-Based Byte-Level Text Classification Without Tokenization or Attention by George Fountzoulas (2026) 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-2604-07969.txt.
Description
Fountzoulas presents Kathleen, a text classification system that works directly on raw UTF-8 bytes with frequency-domain processing. It needs no tokenizer, no attention mechanism, and under 470,000 parameters. Recurrent oscillator banks use damped sine-wave convolutions with time memory for linear-time sequence work. An FFT-Rotate wavetable encoder maps all 256 byte values with a single learnable vector. Phase harmonics add a sine-wave nonlinearity with just six learnable phase values, and further modules add content-dependent memory and sequencing.
Catalog record
| CREATOR | George Fountzoulas |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 24354 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.07969