A mean teacher optimizer that approximates slow natural gradient descent, paired with a new negative log-unlikelihood loss, reduces memorization and privacy leakage on MUSE-News and MUSE-Books, with the strongest variant trading away MMLU.
M., Thakkar, O., and Thakurta, A
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A mean teacher algorithm for unlearning of language models
A mean teacher optimizer that approximates slow natural gradient descent, paired with a new negative log-unlikelihood loss, reduces memorization and privacy leakage on MUSE-News and MUSE-Books, with the strongest variant trading away MMLU.