TOFU - LLaMA 2 1% Forget: leaderboard
TOFU LLaMA 2 1% forget split measuring machine-unlearning quality multiplied by retained model utility.
Metric: Forget Quality x Model Utility. Source: huggingface.co. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | LLaMA 2-7B - Retain Model (WD=0.01) | 0.62 | |
| 2 | LLaMA 2-7B - Pref. Opt. (WD=0.01) | 0.01 | |
| 3 | LLaMA 2-7B - Grad. Ascent (WD=0.01) | 0.01 | |
| 4 | LLaMA 2-7B - KL Min. (WD=0.01) | 0.01 | |
| 5 | LLaMA 2-7B - Grad. Diff. (WD=0.01) | 0 | |
| 6 | LLaMA 2-7B - Finetune Model (WD=0.01) | 0 |
Interactive version: theaggregate.ai/benchmark?slug=tofu-llama-2-1-forget · How It Works · Data refreshed daily, snapshot 2026-10-20.