[KongchangAI]
BenchmarkTask of Fictitious Unlearning

TOFU

A standard benchmark for evaluating machine unlearning, composed of fictional author question-answer pairs, used to test whether a model can forget a specified author while retaining the remaining knowledge.

Timeline (last 90 days)

Nov 17

LEAF-OPD achieved the best results on the TOFU benchmark in Forget Quality, Model Utility and forget ROUGE-L, while maintaining competitive retain performance.

Unverified50%
Nov 17

TOFU, RWKU and MUSE-News evaluate unlearning methods from three dimensions respectively—synthetic data, real people, and real corpora—and combining the three allows cross-validation of method effectiveness.

Unverified50%
Nov 16

Experiments on the TOFU and WMDP benchmarks show that UUE and UUE-L significantly outperform existing methods in unlearning effectiveness.

Unverified50%
Nov 16

TOFU (Task of Fictitious Unlearning) is a standard benchmark for evaluating machine unlearning, consisting of fictitious author question-answer pairs, used to test whether a model can forget a specified author while retaining the rest of its knowledge.

Unverified50%
Oct 14

The TOFU/MOFU/BOFU funnel framework divides search intent into three stages: awareness, evaluation, and decision.

Unverified50%

All Facts (5)

Source Articles