# Three New Papers Attack the Prefill and Key-Value Cache Costs That Make Long Context Expensive

Radius-bounded sparse prefill, learned per-context sparsity during decoding, and attention-informed expert routing all target the same cost constraint from different points in the stack.

- Published: 2026-09-21T06:21:17.793Z
- Canonical: https://polylog.news/ai/2026-09-21/three-new-papers-attack-the-prefill-and-key-value-cache-cost
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [arXiv (RBS-Attention)](https://arxiv.org/abs/2609.20971), [arXiv (Elastic Threshold Attention)](https://arxiv.org/abs/2609.20888), [arXiv (Attention-Aware Routing)](https://arxiv.org/abs/2609.20974)

A cluster of papers posted to arXiv on Monday converges on the same economic problem: long-context inference is now dominated by costs that have nothing to do with how many parameters a model has. RBS-Attention targets prefill, the phase wh…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-09-21/three-new-papers-attack-the-prefill-and-key-value-cache-cost (subscription information: https://polylog.news/pricing).