gemma-2-2b-it-rmu-wmdp-cyber – AI Model by shirasko | AlphaNeural AI | AlphaNeural AI

You can deploy this model and start earning money today!

shirasko/
gemma-2-2b-it-rmu-wmdp-cyber

like

0

text-generation

google/gemma-2-2b-it

text-generation-inference

endpoints_compatible

Views

No views yet

Unlearned Checkpoint

Field	Value
Unlearning method	RMU
Base model	`google/gemma-2-2b-it`
Target concept	wmdp-cyber
Checkpoint type	Full Model Weights
Rank / seed	100 / 42
Train eval protocol	mc

Unlearning Configuration

Selected hyperparameters (from unlearned_checkpoints.json):

Parameter	Value
`alpha`	50
`delta_embed`	0
`k_features_embed`	0
`layer_id`	8
`layer_ids`	6,7,8
`lr`	0.0003
`n_tokens_edited`	0
`param_ids`	6
`setting_name`	S2_lid8_L678
`steering`	300

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

Headline scores used for checkpoint selection:

Metric	Train (after unlearning)	Test (after unlearning)
Efficacy	1	0
Specificity	0.962	0.98
Harmonic mean	0.98	0
Relearning QA (MC)	—	0.24

Full Evaluation (baseline → unlearned)

From evaluation/score_comparison.csv:

Metric	Baseline (train)	After unlearn (train)	Baseline (test)	After unlearn (test)
QA accuracy	0.3	0.24	0.3	0.32
QA fraction	1	0	1	1
SimDom accuracy	0.44	0.48	0.36	0.38
SimDom fraction	1	1	1	1
MMLU accuracy	0.52	0.5	0.551	0.539
MMLU fraction	1	0.926	1	0.96

Files in This Repository

File	Description
`unlearned_checkpoints.json`	Checkpoint metadata & hyperparameters
`evaluation/evaluation_summary.json`	Full evaluation payload (train/test/relearning)
`evaluation/score_comparison.csv`	Baseline vs. unlearned comparison table