Unverified50% confidenceFactExact time
BERT只使用Transformer的编码器专练理解，GPT只使用解码器专练生成
1
Sources
50%
Confidence
Long-term
Relevance
7/12/2026
First Seen
Sources
CNN到Transformer：AI理解人类语言的进化之路
bilibiliAI-Agent搭建7/9/2026
Related Claims
UnverifiedGPT-2采用纯解码器的Transformer堆叠结构，去掉了编码器-解码器的交叉注意力部分，专注于单向自回归语言建模72% similarPartially VerifiedTransformers come in three variants: encoder-only (like BERT), decoder-only (like GPT), and encoder-decoder (like T5)71% similarUnverified现代LLM普遍采用的GPT架构本质上是仅使用Transformer解码器（Decoder-only）部分的自回归语言模型69% similarUnverifiedTransformer架构分化为BERT类（Encoder-only，适合理解任务）和GPT类（Decoder-only，适合生成任务）两大路线68% similarVerifiedBERT 采用编码器架构，GPT 系列采用解码器架构，T5 和 BART 采用编码器-解码器架构68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/490824API
curl https://kongchang.com/api/v1/knowledge/claims/490824MCP
get_claim(id=490824)