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

Unverified70% confidenceFactExact time

Top-P核采样（Nucleus Sampling）由Holtzman等人于2020年提出，通过动态截取累积概率达到阈值P的最小词汇子集进行采样

2

Sources

70%

Confidence

Long-term

Relevance

7/14/2026

First Seen

Sources

Fable延期至7月19日：AI交互叙事产品的挑战与启示

Related Claims

UnverifiedTop-p（核采样）通过限定候选token的累积概率范围来控制输出多样性，与Temperature配合使用可实现更精细的输出调节70% similar UnverifiedCoze的Top-P核采样机制是从最高概率的词开始累加，直到累积概率达到设定的P值，只从这些词中采样67% similar UnverifiedTop-p（核采样）通过动态截断概率质量限制候选词范围，相比固定截断 Top-k 具有更好的自适应性64% similar UnverifiedDMARC建议采用渐进式部署策略：先设置p=none监控收集数据，观察2-4周后逐步收紧到p=quarantine乃至p=reject63% similar UnverifiedGPTQ量化算法由Tim Dettmers等人于2023年提出，通过二阶梯度信息逐层优化量化误差62% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/503408

API

curl https://kongchang.com/api/v1/knowledge/claims/503408

MCP

get_claim(id=503408)