Unverified50% confidenceFactExact time
The generation pipeline of Diffusion-LM consists of three steps: mapping words to continuous vectors using learnable word embeddings, iteratively denoising via the diffusion model to obtain a sequence of continuous latent variables, and restoring continuous vectors to discrete text through a Rounding operation.
1
Sources
50%
Confidence
Long-term
Relevance
9/2/2026
First Seen
Sources
Related Entities
Related Claims
Unverified离散扩散语言模型主流方案包括：将token映射到连续嵌入空间进行扩散（如Diffusion-LM），或设计离散扩散过程用掩码模拟加噪（如D3PM、MDLM）74% similarUnverifiedDiffusion-LM uses learnable embeddings rather than fixed embeddings because the geometric structure of the embedding space needs to mutually adapt to the denoising dynamics of the diffusion process.73% similarUnverifiedStable Diffusion 基于去噪扩散概率模型（DDPM）理论，将扩散过程从像素空间转移到低维潜在空间，由VAE负责编解码，CLIP模型将文本提示编码为语义向量68% similarUnverifiedLyra breaks down the prompt generation process into four phases: Deconstruct, Diagnose, Develop, and Deliver.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/843367API
curl https://kongchang.com/api/v1/knowledge/claims/843367MCP
get_claim(id=843367)