The problem of model collapse: why recursive training might lead to the decline of large language models

Herasimovich, I. I.

Full metadata record

DC Field	Value	Language
dc.contributor.author	Herasimovich, I. I.	-
dc.coverage.spatial	Минск	en_US
dc.date.accessioned	2026-08-20T06:54:35Z	-
dc.date.available	2026-08-20T06:54:35Z	-
dc.date.issued	2026	-
dc.identifier.citation	Herasimovich, I. I. The problem of model collapse: why recursive training might lead to the decline of large language models / I. I. Herasimovich // Актуальные вопросы экономики и информационных технологий : сборник материалов докладов 62-й научной конференции аспирантов, магистрантов и студентов БГУИР, Минск, 13–17 апреля 2026 г. / Белорусский государственный университет информатики и радиоэлектроники. – Минск, 2026. – С. 725–726.	en_US
dc.identifier.uri	https://libeldoc.bsuir.by/handle/123456789/65575	-
dc.description.abstract	The paper examines the phenomenon of model collapse in Large Language Models caused by recursive training on synthetic data. It analyses the statistical mechanics of distributional decay, identifies the progressive stages of knowledge degradation, evaluates the sociological impact on digital platforms, and proposes technical mitigation strategies to preserve the integrity of the artificial intelligence ecosystem.	en_US
dc.language.iso	en	en_US
dc.publisher	БГУИР	en_US
dc.subject	материалы конференций	en_US
dc.subject	recursive learning	en_US
dc.subject	large language models	en_US
dc.subject	decreased efficiency	en_US
dc.subject	model training	en_US
dc.subject	artificial intelligence	en_US
dc.subject	neural network training	en_US
dc.subject	model degradation	en_US
dc.subject	scaling issues	en_US
dc.title	The problem of model collapse: why recursive training might lead to the decline of large language models	en_US
dc.type	Article	en_US
Appears in Collections:	Актуальные вопросы экономики и информационных технологий : материалы 62-й научной конференции аспирантов, магистрантов и студентов (2026)