A fonte Google News publicou a seguinte informação sobre Camada de memória salva estado KV de modelos Gemma 4 e carrega janela de 50 milhões de tokens sem recomputar:
> [2610.10845] Real Long-Term Memory for AI: A 50-Million-Token Window That Is Faster and Cheaper Than Recompute Skip to main content Search arXiv Press Enter to search · Advanced search -- Computer Science Computation and Language arXiv:2610.10845 (cs) [Submitted on 7 Oct
> 2026] Title: Real Long-Term Memory for AI: A 50-Million-Token Window That Is Faster and Cheaper Than Recompute Authors: Sietse Schelpe View a PDF of the paper titled Real Long-Term Memory for AI: A 50-Million-Token Window That Is Faster and Cheaper Than Recompute, by Sietse
> Schelpe View PDF HTML (experimental) Abstract: A large language model can only use the text that fits in its context window, and it recomputes its internal key-value (KV) state for a prompt every time the prompt is sent.
Evidências disponíveis
Os trechos acima foram preservados literalmente da fonte consultada.
