Model collapse – how LLMs become worse when trained on their own output (ibm.com) 2 points by daymos 8mo ago ↗ HN
2 comments
[ 18.1 ms ] story [ 231 ms ] thread