Unverified60% confidenceFactExact time
Sora采用了Diffusion Transformer（DiT）架构，将Transformer全局注意力机制引入扩散框架
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/12/2026
First Seen
Valid until: 10/10/2026
Sources
AI生成视频的进步与瓶颈：光影惊艳，逻辑存疑
redditr/midjourney7/11/2026
Related Claims
Partially VerifiedSora采用DiT（Diffusion Transformer）架构，将视频压缩为时空patch序列后输入Transformer，以Transformer替代传统U-Net作为去噪骨干网络79% similarVerifiedFLUX采用了DiT（Diffusion Transformer）架构，用纯Transformer替代了传统的U-Net结构70% similarUnverifiedTransformers库集成了GPTQ、AWQ、bitsandbytes等量化技术和Flash Attention加速方案70% similarUnverifiedFlux 是基于扩散 Transformer（DiT）架构的新一代模型68% similarUnverifiedMegatron-LM最早系统化提出了针对Transformer的张量并行方案，将注意力头和FFN层分别切分到不同设备67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/494208API
curl https://kongchang.com/api/v1/knowledge/claims/494208MCP
get_claim(id=494208)