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% confidenceOpinionExact time

顶级会议的马太效应产生自我强化的正反馈循环，形成赢家通吃格局

1

Sources

50%

Confidence

Long-term

Relevance

7/16/2026

First Seen

Sources

AI学术会议生态过度集中：我们失去了什么？

Related Claims

Unverified用户在面对海量选择时倾向于选择已有社会证明的项目，形成强者愈强的累积优势正反馈循环70% similar Unverified元认知反馈的奖励机制：对正确答案高置信度给予奖励，对错误答案高置信度给予惩罚，对不确定问题恰当表达犹豫也予以鼓励68% similar Unverified元认知反馈通过奖励置信度与准确率的一致性，在奖励信号中引入认知诚实度这一与内容质量正交的新维度67% similar Unverified奖励塑形（Reward Shaping）通过设计中间奖励引导学习方向，如将对Boss造成每点伤害设为中间奖励67% similar UnverifiedGao et al.（2023）的研究发现随着策略模型对奖励模型的优化程度增加，真实人类偏好评分先升后降呈现倒U形曲线66% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/527917

API

curl https://kongchang.com/api/v1/knowledge/claims/527917

MCP

get_claim(id=527917)