Unverified50% confidenceOpinionExact time
DiffusionGemma 大概率采用掩码扩散方案，前向过程逐步将 token 替换为 [MASK] 标记
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/22/2026
First Seen
Valid until: 11/20/2026
Sources
Related Entities
Related Claims
UnverifiedMDLM（Sahoo et al., 2024）专注于吸收态扩散，前向过程只将 token 替换为 [MASK]66% similarUnverifiedD3PM（Austin et al., 2021）将扩散过程推广到离散状态空间，通过转移矩阵描述 token 在离散状态间的随机跳转60% similarVerifiedStable Diffusion compresses the diffusion process from pixel space to latent space, implemented by a Variational Autoencoder (VAE).59% similarUnverified投机解码通过小模型预测候选Token、大模型批量验证提升吞吐量；量化技术将权重从FP16压缩至INT8/INT4降低显存占用；MoE架构通过稀疏激活降低单次推理计算量58% similarUnverified强行取极值样本会引入离群点，迫使SVM构建更复杂的决策边界，高阶poly核在这种数据上收敛更慢56% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/787219API
curl https://kongchang.com/api/v1/knowledge/claims/787219MCP
get_claim(id=787219)