Unverified50% confidenceFactExact time

Gemma系列使用的SentencePiece分词器词表约含256,000个条目

1

Sources

50%

Confidence

Long-term

Relevance

7/14/2026

First Seen

Sources

GDScript+Vulkan在Godot中运行Gemma大语言模型实验

Related Claims

UnverifiedGemma系列使用SentencePiece分词器，在BPE基础上引入Unigram语言模型，词表约含256,000个条目85% similar Unverified网文平台单章通常在2000-4000字之间57% similar Unverified牛津词典收录约17万英语词汇57% similar Unverified字库覆盖建议：启蒙阶段300字级足够（覆盖日常口语高频字），学前衔接选500-800字级，对应小学一年级识字量约700字54% similar Unverified标准自回归语言模型的词表通常包含32,000到128,000个token53% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/504403

API

curl https://kongchang.com/api/v1/knowledge/claims/504403

MCP

get_claim(id=504403)