# Small-model papers test linear attention, integer number formats and new sequence models for compute-limited training

Kairos pairs bidirectional gated DeltaNet with sliding-window attention for a byte-level multimodal model trained on one machine, and a second paper proposes algebraic integer formats that reduce inference to integer arithmetic.

- Published: 2026-10-12T06:16:38.065Z
- Canonical: https://polylog.news/ai/2026-10-12/small-model-papers-test-linear-attention-integer-number-form
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv cs.LG (Kairos)](https://arxiv.org/abs/2610.12495), [arXiv cs.LG (Algebraic Integer Quantization)](https://arxiv.org/abs/2610.12494), [arXiv cs.LG (Resonant Sequence Models)](https://arxiv.org/abs/2610.12496), [Polylog editors](https://polylog.news)

"Kairos: Toward a Byte-Level Foundation Model for Every Modality" starts from a constraint most labs do not face. A small language model trained on a single local machine leaves little budget for experimenting with architectures. Kairos com…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-10-12/small-model-papers-test-linear-attention-integer-number-form (subscription information: https://polylog.news/pricing).