https://jaehun.me/en/tags/test-time-training/Technical notes by Jaehun Ryu, a deep learning compiler engineer, on LLM inference, KV caches, compilers, C++, AI papers, and open-source code.en-UShttps://jaehun.me/en/posts/paper-2611-17331/https://jaehun.me/en/posts/paper-2611-17331/Sun, 15 Nov 2026 00:00:00 +0900<p><a href="https://arxiv.org/abs/2611.17331"target="_blank" class="inline-flex items-center gap-1" >Paper<svg class="h-3 w-3 flex-shrink-0" id="external-link" xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24"><path fill="none" stroke="currentColor" stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M15 3h6v6m-11 5L21 3m-3 10v6a2 2 0 0 1-2 2H5a2 2 0 0 1-2-2V8a2 2 0 0 1 2-2h6"/></svg> </a></p> <h2 id="1-introduction-every-new-view-reads-the-entire-cache">1. Introduction: every new view reads the entire cache<a href="#1-introduction-every-new-view-reads-the-entire-cache" class="heading-anchor" aria-label="Link to this section"></a></h2><p>Among models that generate video from multiple photographs at a new angle, the highest-quality design stores scene information in a Transformer&rsquo;s key-value cache. The price of this design is that every new view requires re-reading the entire cache, so rendering slows down as the number of input photographs grows. This paper proposes a method that bakes that information once into the weights of a small neural network, making the cost of rendering a new view independent of the number of input photographs.</p>