English(EN) CASIA's PhiZero Gives World Models a 'Physical Language' and Cuts Tokens 175x

CASIA 的 PhiZero 引入世界模型的“物理语言”，将 Token 削减 175 倍

作者 PulseAugur 编辑部 · [1 个来源] · 2026-08-07 08:37

中国科学院（CASIA）的研究人员开发了 PhiZero，这是一种新颖的世界模型，它利用学习到的离散“物理语言”进行视频生成。这种创新的方法显著减少了表示视频内容所需的 Token 数量，将一个四秒钟的剪辑的 Token 数量减少了 175 倍。 AI

影响这种方法可能导致 AI 模型中更高效的视频生成和表示。

排序理由研究论文发布，详细介绍了新模型和方法论。[lever_c_demoted from research: ic=1 ai=1.0]

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 研究论文发布，详细介绍了新模型和方法论。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: model release, paper; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 56 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

Pandaily TIER_1 English(EN) · [email protected] (Pandaily) · 2026-08-07 08:37

中科院PhiZero赋予世界模型“物理语言”，将Token削减175倍

Researchers at the Institute of Automation under the Chinese Academy of Sciences (CASIA) published PhiZero, a world model that reasons in a learned discrete 'physical language' before rendering video. The approach collapses the number of tokens needed to represent a four-second c…