跳转至

模型坍塌与应对策略综述

文章背景与核心概要

随着生成式AI（GenAI）的迅猛发展，从业者越来越多地使用AI合成的数据来训练新一代模型，以应对激增的数据需求。然而，这种自我消耗的循环带来了一个关键漏洞：模型坍塌（Model Collapse, MC），即在合成数据上进行迭代训练最终会导致模型性能退化和可信度丧失。

本文对模型坍塌现象进行了首个全面且最新的综述。文章总结了各个应用场景中的最新进展，审查了旨在缓解模型坍塌的现有对策，并概述了尚未解决的研究挑战与未来的发展机遇。

核心内容回顾与对策研究

Summary

As Generative AI (GenAI) advances rapidly, practitioners increasingly train next-generation models on AI-synthesized data to meet surging data demands. However, this self-consuming cycle introduces a critical vulnerability: Model Collapse (MC), where iterative training on synthetic data eventually leads to model degradation and loss of trustworthiness.

This paper provides the first comprehensive, up-to-date overview of the Model Collapse phenomenon. It consolidates progress across various application scenarios, examines existing countermeasures designed to mitigate MC, and outlines open research challenges and future opportunities.

文档元数据

Document Metadata

字段 详情
标题 Reviewing Model Collapse and Countermeasures
作者 Xihao Xie, Beichen Hu
提交日期 2026年6月17日
一级学科 人工智能 (cs.AI)
二级学科 机器学习 (cs.LG)
arXiv 标识符 arXiv:2608.21366 [cs.AI]
DOI 10.48550/arXiv.2608.21366
期刊引用 Proceedings of the IEEE International Conference on Advances in Artificial Intelligence and Machine Learning (AAIML), 2026 (11页, 1幅图)

摘要

Driven by massive amounts of web-scale data, generative AI (GenAI) has achieved remarkable progress, enabling various applications in diverse sectors. The advances of GenAI have actuated practitioners to use AI-synthesized data for training next-generation AI models. Undeniably, using synthetic data has alleviated the increasing stringent demand for data supply. Unfortunately, it also introduces a new critical issue: in a self-consuming cycle between model and data, the model ultimately collapse, raising more trustworthiness concerns to GenAI. In recent years, increasingly more studies have investigated the phenomenon of model collapse (MC) and explored potential solutions to mitigate it. However, the review of the phenomenon of MC still remains blank. To fill this gap, this paper provides an up-to-date overview of these studies for consolidating and reviewing the progress of MC in different application scenarios and countermeasures for mitigating MC. We also highlight challenges and future research opportunities.

访问与全文链接

Access & Full-Text Links

外部资源与引用

External Resources & Citations