[Deep Dive] Breaking Down the Inference Costs of Input-Heavy AI Agents ~Where DeepSeek-V4.1-Flash Cut Costs~

Memorizing SideI recently read the DeepSeek-AI paper, "DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression", ...

Why place CloudFront in front of S3

As part of understanding our company's AWS configuration, I had the opportunity to explain internally "why we place ...
BIOENGINEER

Merging Memory: New KVCB Method Slims AI Caches Without Losing the Long Tail

Researchers at Anhui University have developed KVCB, a key-value cache merging method that compresses large language model ...
Science

Contagion-preserving compression for multiscale epidemic modeling

Understanding how contagion unfolds on large, heterogeneous networks is essential for predicting and controlling spreading ...
blog.csdn

DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression

论文：DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression（arXiv:2609.19969，DeepSeek-AI，2026-09-17） 一句话：DeepSeek 又整了个"胃口很小、脑子很大"的模型——552B 的 MoE 大模型，能吃100 万 token 的上下文 ...