LLMs corrupt your documents when you delegate(arxiv.org)
454 points by rbanffy | 177 comments
arXiv paper introducing DELEGATE-52, a benchmark showing LLMs corrupt about 25% of document content during long delegated editing workflows.
1 story tagged delegate-52, newest first.
454 points by rbanffy | 177 comments
arXiv paper introducing DELEGATE-52, a benchmark showing LLMs corrupt about 25% of document content during long delegated editing workflows.