tofu_Llama-3.1-8B-Instruct_forget10_SatImp
open-unlearning/tofu_Llama-3.1-8B-Instruct_full unlearned on the TOFU forget10 split with SatImp, trained with the open-unlearning framework. Used as a weight-unlearning baseline / draft model in the Speculative-Decoding-Unlearning project.
Full training config: .hydra/config.yaml. TOFU evaluation outputs: evals/.
Method hyperparameters
gamma: 1.0
alpha: 0.1
retain_loss_type: NLL
beta1: 5.0
beta2: 1.0
TOFU summary metrics
| metric | value |
|---|---|
| exact_memorization | 0.5034 |
| extraction_strength | 0.0538 |
| forget_Q_A_PARA_Prob | 0.0063 |
| forget_Q_A_gibberish | 0.6780 |
| forget_quality | 0.0000 |
| forget_truth_ratio | 0.8045 |
| mia_loss | 0.0280 |
| mia_min_k | 0.2373 |
| mia_min_k_plus_plus | 0.9972 |
| mia_zlib | 0.0192 |
| model_utility | 0.6151 |
| privleak | 23.0464 |