Unverified50% confidenceFactExact time
SentencePiece由Google于2018年推出，支持无监督训练，无需预先进行空格分词，Google的T5、LLaMA等模型均采用其Unigram语言模型算法
1
Sources
50%
Confidence
Long-term
Relevance
7/25/2026
First Seen
Sources
AI工业时代来临：算力争夺、芯片重构与GPT-5.6作弊风波全解析
bilibili陈工AI观测7/5/2026
Related Claims
UnverifiedSentencePiece was developed by Google and learns subword units directly from raw text without requiring a pre-tokenization step.81% similarUnverifiedGoogle于2022年发布SayCan，让语言模型仅负责生成可行动作序列，具体执行交由预训练底层技能策略完成76% similarUnverifiedBERT was released by Google in 2018, employing bidirectional pre-training that randomly masks 15% of words via a masked language model and requires their restoration.73% similarUnverifiedGoogle Dialogflow使用了BERT等预训练模型进行意图分类和槽位填充71% similarUnverifiedGoogle released BERT in 2018, marking the beginning of the pre-training era.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/613125API
curl https://kongchang.com/api/v1/knowledge/claims/613125MCP
get_claim(id=613125)