Unverified50% confidenceFactExact time
BERT（2018年，Google）采用掩码语言模型（MLM）实现双向上下文理解，但无法生成连贯文本
1
Sources
50%
Confidence
Long-term
Relevance
7/6/2026
First Seen
Sources
GLM-4系列开源逆袭：AI战争蔓延至内存争夺与太空算力
bilibili智灵Shaw7/3/2026
Related Claims
UnverifiedGoogle DeepMind的研究表明LLM可作为翻译器将非形式化数学语句转译为Lean或Isabelle表达，但准确率仍远未达到可部署水平69% similarUnverifiedLLM通过在海量文本语料上进行无监督预训练（预测下一个词）来习得语言能力67% similarUnverified2006年Google翻译初始系统使用了一个在数万亿词文本上训练的5-gram语言模型，这很可能是世界上第一次万亿token规模的语言模型训练66% similarVerified2022年谷歌Brain团队在论文《Emergent Abilities of Large Language Models》中首次系统记录了大语言模型的'涌现能力'现象64% similarPartially VerifiedLaMDA (Language Model for Dialogue Applications) was Google's language model optimized for open-domain dialogue, trained on the Transformer architecture and capable of multi-turn conversations63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/115228API
curl https://kongchang.com/api/v1/knowledge/claims/115228MCP
get_claim(id=115228)