Unverified50% confidenceFactExact time
The GPT series uses only the decoder portion and focuses on text generation, while BERT uses only the encoder portion and excels at text understanding.
1
Sources
50%
Confidence
Long-term
Relevance
9/8/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedBERT只使用Transformer的编码器专练理解，GPT只使用解码器专练生成76% similarUnverifiedGPT-3 的预训练语料以自然语言为主，代码仅占约 1%，而 Codex 的微调数据集中代码占比超过 90%74% similarUnverifiedGPT-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% similarUnverifiedModern large language models (such as the GPT series) almost universally adopt a decoder-only architecture, using masked self-attention to ensure that autoregressive generation does not leak future token information.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/879892API
curl https://kongchang.com/api/v1/knowledge/claims/879892MCP
get_claim(id=879892)