LLM Unlearning on TOFU (forget01) benchmark leaderboard · SOTA2 Research

LLM Unlearning on TOFU (forget01)

1FQRetained LLM

Evaluation Results

Method	Links
Retained LLM2026.06		1	0.6267	0.408
RULE_GRPO2026.06		0.165	0.6151	0.1322
ReRULE_GRPO2026.06		0.165	0.6076	0.1322
GA/SGA2026.06		0.0068	0.599	0.4817
GD2026.06		0.0068	0.5998	0.4869
DPO-RT2026.06		0.0068	0.6322	0.2595
SGA2026.06		0.0068	0.6027	0.4792
KL2026.06		0.003	0.5994	0.4922
LLMU2026.06		0.003	0.5999	0.4891
PO2026.06		0.003	0.6323	0.1752
NPO-RT2026.06		0.003	0.5994	0.5049
FLAT2026.06		0.003	0.6304	0.4825
Original LLM2026.06		0	0.6239	0.9851