CatShift detects training data membership in LLMs by comparing output shifts induced by fine-tuning on member versus non-member data, relying on catastrophic forgetting without requiring logit access.
You only query once: an efficient label-only membership inference attack
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Hey, That's My Data! Token-Only Dataset Inference in Large Language Models
CatShift detects training data membership in LLMs by comparing output shifts induced by fine-tuning on member versus non-member data, relying on catastrophic forgetting without requiring logit access.