<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>写作 · Writing</title><description>关于 AI、气候科学与流式多模态模型的一些思考。Notes on AI, climate science and streaming multimodal models.</description><link>https://jialiangz.github.io/writing/</link><item><title>流式推理中的思维链：为什么要边看边想</title><link>https://jialiangz.github.io/writing/posts/streaming-mllm-reasoning/</link><guid isPermaLink="true">https://jialiangz.github.io/writing/posts/streaming-mllm-reasoning/</guid><description>在视频理解这类流式任务上，模型如果必须等完整输入才能开始推理，就会失去实时性。这篇文章梳理我们在 Streaming CoT 上的几个观察。</description><pubDate>Sun, 18 Oct 2026 00:00:00 GMT</pubDate></item><item><title>Chain-of-Thought in Streaming Reasoning: Think While You Look</title><link>https://jialiangz.github.io/writing/posts/streaming-mllm-reasoning-en/</link><guid isPermaLink="true">https://jialiangz.github.io/writing/posts/streaming-mllm-reasoning-en/</guid><description>In streaming video understanding, a model that must wait for the full input before reasoning loses real-timeness. Here are a few observations from our work on Streaming CoT.</description><pubDate>Sun, 18 Oct 2026 00:00:00 GMT</pubDate></item><item><title>北极海冰的逐日预测，难点不在模型</title><link>https://jialiangz.github.io/writing/posts/sea-ice-daily-forecast/</link><guid isPermaLink="true">https://jialiangz.github.io/writing/posts/sea-ice-daily-forecast/</guid><description>做了一段时间的逐日海冰浓度预测后，我越来越确信：真正的瓶颈是数据对齐，而不是网络结构。</description><pubDate>Wed, 02 Sep 2026 00:00:00 GMT</pubDate></item><item><title>A site for writing</title><link>https://jialiangz.github.io/writing/posts/a-site-for-writing/</link><guid isPermaLink="true">https://jialiangz.github.io/writing/posts/a-site-for-writing/</guid><description>Why I moved my notes out of a private notebook and into a static site: versioning, multilingual by default, and no lock-in.</description><pubDate>Fri, 21 Aug 2026 00:00:00 GMT</pubDate></item></channel></rss>