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

2023年谷歌DeepMind等机构的研究证实，模型的记忆化（Memorization）行为与训练数据的重复次数密切相关

1

Sources

50%

Confidence

Long-term

Relevance

7/12/2026

First Seen

Sources

纽约时报指控OpenAI隐匿证据，ChatGPT版权诉讼再升级

rss7/9/2026

Related Claims

Unverified卡内基梅隆大学等机构研究证明大型扩散模型存在「训练数据记忆」现象，可在特定提示词下高度还原训练集中的版权图像77% similar UnverifiedSWE-bench上已有多个研究团队记录到「patch记忆」现象，即模型通过识别问题描述特征检索训练时见过的补丁71% similar Unverified自适应推理技术路线包括难度感知路由、早停机制、预算控制训练及Mixture-of-Depths等，Meta、Google DeepMind等机构在2024-2025年间发表了相关工作71% similar Unverified主流基准测试题目在互联网上已被广泛讨论，模型训练数据中可能存在对题目的记忆，导致基准分数高估模型在真实未见任务上的能力70% similar Verified涌现能力指模型在规模达到一定阈值后突然表现出训练阶段未被明确设计的能力，这一现象最早由Google Research团队在2022年系统性地描述67% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/494400

API

curl https://kongchang.com/api/v1/knowledge/claims/494400

MCP

get_claim(id=494400)