Unverified60% confidenceFactExact time

Text generation in large language models is autoregressive, and the number of output tokens is nearly proportional to response time

2

Sources

60%

Confidence

Long-term

Relevance

9/1/2026

First Seen

Sources

GLM-5.3 Flash's Overly Verbose Thinking Process? Token Costs of Reasoning Models Spark Heated Debate

Related Entities

Large Language Modelconcept

Related Claims

UnverifiedThe output of modern large language models is a streaming process generated token by token.77% similar Verified大语言模型通过在数千亿乃至数万亿个词元（Token）的文本数据上进行自监督预训练76% similar Unverified当训练数据中某些文本出现频率极高时，大语言模型可能产生记忆化（Memorization）现象，能够近乎逐字输出特定段落76% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/838003

API

curl https://kongchang.com/api/v1/knowledge/claims/838003

MCP

get_claim(id=838003)