CoolFace
Modelpublic

shirasko/llama-3.1-8b-instruct-rmu-wmdp-bio

sourceHugging Faceupdated 2mo agoView on Hugging Face
0likes15downloads
Model Card

Unlearned Checkpoint

FieldValue
Unlearning methodRMU
Base modelmeta-llama/Llama-3.1-8B-Instruct
Target conceptwmdp-bio
Checkpoint typeFull Model Weights
Rank / seed200 / 42
Train eval protocolmc

Unlearning Configuration

Selected hyperparameters (from unlearned_checkpoints.json):

ParameterValue
alpha50
delta_embed0
k_features_embed0
layer_id7
layer_ids5,6,7
lr0.0001
n_tokens_edited0
param_ids6
setting_nameS1lid7L567
steering30

Primary Unlearning Metrics (held-out test, MC protocol)

Headline scores used for checkpoint selection:

MetricTrain (after unlearning)**Test (after unlearning)**
Efficacy0.7171
Specificity0.8490.749
Harmonic mean0.7770.856
Relearning QA (MC)—0.7

Full Evaluation (baseline → unlearned)

From evaluation/score_comparison.csv:

MetricBaseline (train)After unlearn (train)Baseline (test)**After unlearn (test)**
QA accuracy0.780.40.780.22
QA fraction10.28310
SimDom accuracy0.820.740.90.68
SimDom fraction10.8610.662
MMLU accuracy0.620.560.650.595
MMLU fraction10.83810.862

Files in This Repository

FileDescription
unlearned_checkpoints.jsonCheckpoint metadata & hyperparameters
evaluation/evaluation_summary.jsonFull evaluation payload (train/test/relearning)
evaluation/score_comparison.csvBaseline vs. unlearned comparison table