Unmasking the Secret Code: Using Canaries to Show How Masked Context Controls Memorization in SFT

arXiv:2610.20656v1 Announce Type: new Abstract: Canaries, or synthetic training records containing known secrets, are widely used to assess whether language models reveal information from their training data. But how much do these measurements depend on the canaries themselves? We study this question through controlled fine-tuning experiments with language and vision-language models, spanning synthetic tasks and tasks where training improves useful performance.

Read the full article at arXiv cs.CR →