Unverified50% confidenceFactExact time
模式坍缩（Mode Collapse）最早在GAN领域被系统性描述，指生成器陷入只生成少数几种高置信度样本的局面
1
Sources
50%
Confidence
Long-term
Relevance
7/14/2026
First Seen
Sources
口语化采样登顶ICML：提示工程配得上顶会吗？
Related Claims
UnverifiedGAN training faces the challenge of mode collapse, where the generator may fall into a local optimum of generating only a few highly similar samples78% similarVerifiedGANs have inherent drawbacks such as training instability and susceptibility to mode collapse, which prompted the research community to shift toward diffusion models that generate images through a gradual denoising Markov chain process.75% similarUnverifiedGAN训练中常见的模式崩溃（Mode Collapse）现象可能导致生成数据多样性严重不足71% similarUnverifiedGANs存在训练不稳定和模式崩溃（生成内容多样性不足）等固有局限65% similarUnverifiedModel collapse occurs in two progressive stages: in the early collapse stage the output mean drifts, and in the late collapse stage the output variance approaches zero and generated content converges into a homogeneous, singular pattern.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/504319API
curl https://kongchang.com/api/v1/knowledge/claims/504319MCP
get_claim(id=504319)