Model collapse – how LLMs become worse when trained on their own output (ibm.com) 2 pts|11 days ago|1 comment
