# Decode-Latency Feedback Prefill: A Model-Free Controller and Its Generalization Limits

> Полный текст доступен на английском

**Published:** 1 октября 2026 г.
**Author:** Gaurav Agarwal, Ashish Garg, Isha Singhal
**Source:** arXiv — cs.AI
**Language:** en-US
**HTML:** https://aiqazaqstan.com/ru/articles/decode-latency-feedback-prefill-a-model-free-controller-and-its-generali

Concurrent autoregressive inference creates a fundamental interference problem: prefilling a newly arrived long prompt can delay tokens for requests that are already decoding. Fixed prefill chunks reduce this interference, but the best chunk size depends on the model, hardware,…

> Материал агрегирован из первоисточника: **arXiv — cs.AI**
> [Читать полностью в arXiv — cs.AI](https://arxiv.org/abs/2609.38386)

**Теги:** #arXiv

---

*AI Qazaqstan — non-commercial informational portal about AI in Kazakhstan. Machine translations and summaries are AI-generated.*