Unverified50% confidenceFactExact time
研究者提出了掩码扩散（Masked Diffusion）、吸收态扩散（Absorbing Diffusion）等变体以解决文本离散空间的扩散挑战
1
Sources
50%
Confidence
Long-term
Relevance
7/8/2026
First Seen
Sources
Google's AI Blitz: Gemini 3.5 Live Translate, NotebookLM Upgrades, and DiffusionGemma Explained
twitterGoogleAI6/12/2026
Related Claims
Unverified文本扩散采用掩码扩散（Masked Diffusion）方案，用[MASK]token替代真实token模拟损坏状态，再训练模型恢复原始文本，以块为单位并行输出66% similarUnverified早期的离散扩散语言模型如 MDLM、SEDD 通过在 Token 嵌入空间或掩码空间中定义扩散过程来解决文本离散性挑战66% similarVerifiedAnthropic's 2023 monosemanticity research explored how to separate aliased features using sparse autoencoders.62% similarUnverifiedThe underlying architecture of mainstream text-to-image models such as Stable Diffusion, Midjourney, and DALL-E typically combines a diffusion model with a CLIP text encoder.62% similarVerifiedStable Diffusion's text guidance converts prompts into vectors through the CLIP text encoder.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/238232API
curl https://kongchang.com/api/v1/knowledge/claims/238232MCP
get_claim(id=238232)