Feed Entities Podcasts Deep Dives Developers About

[KongchangAI]

Real-time AI knowledge base. Every claim is traceable to its source, with data freshness under 24 hours.

Powered by 240+ first-hand AI sources

Knowledge

Feed
Hot Entities
Podcasts

Developers

CLI
MCP Server
REST API
About

© 2026 KongchangAI. The freshest, most accurate AI knowledge

Unverified50% confidenceFactExact time

以Transformer架构为基础的语言模型，其推理计算量与参数规模大致呈正比，参数越多单次推理所需的FLOPs越高

1

Sources

50%

Confidence

Long-term

Relevance

7/12/2026

First Seen

Sources

GPT-5.6与Grok 4.5同期发布，Meta推出Agent图像模型

bilibiliEverAI酱7/8/2026

Related Claims

Unverified在传统Transformer架构中，每个token的计算量是固定的，而思维链（Chain-of-Thought）技术通过让模型生成中间推理步骤来增加有效计算量77% similar UnverifiedTransformer架构中模型参数量与推理延迟之间呈近似线性关系76% similar Unverified大语言模型基于Transformer架构，其输出本质上是概率采样过程，由Temperature参数控制采样随机性75% similar Verified大语言模型基于Transformer架构，每次推理都是独立的前向传播过程，不保留任何会话状态75% similar Unverified大语言模型输出的非确定性源于 Transformer 解码阶段的随机采样机制，采样温度、Top-P 等参数决定每次生成内容可能不同74% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/490205

API

curl https://kongchang.com/api/v1/knowledge/claims/490205

MCP

get_claim(id=490205)