Prompt-based fine-tuning (prefix, prompt, P-tuning) shows consistently lower membership inference AUC than parameter-based fine-tuning (full, head, LoRA) across GPT-2, LLaMA-2-7B, and LLaMA-3-1B.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Impact of Fine-Tuning Methods on Memorization in Large Language Models
Prompt-based fine-tuning (prefix, prompt, P-tuning) shows consistently lower membership inference AUC than parameter-based fine-tuning (full, head, LoRA) across GPT-2, LLaMA-2-7B, and LLaMA-3-1B.