llama-3.1-8b-instruct-snmf-wmdp-bio – AI Model by shirasko | AlphaNeural AI

You can deploy this model and start earning money today!

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

like

0

text-generation

llama-3.1-8b-instruct

meta-llama/Llama-3.1-8B-Instruct

text-generation-inference

endpoints_compatible

Views

No views yet

Unlearned Checkpoint

Field	Value
Unlearning method	SNMF
Base model	`meta-llama/Llama-3.1-8B-Instruct`
Target concept	wmdp-bio
Checkpoint type	Full Model Weights
Rank / seed	200 / 42
Train eval protocol	mc

Unlearning Configuration

Selected hyperparameters (from unlearned_checkpoints.json):

Parameter	Value
`coverage_thresh`	0.95
`delta_embed`	0
`delta_in`	4
`delta_out`	1
`feature_source`	all
`k_features_embed`	0
`k_features_mlp_in`	416
`k_features_mlp_out`	330
`layer_hi_in`	16
`layer_hi_out`	10
`layer_lo_in`	0
`layer_lo_out`	0
`n_tokens_edited`	0
`ratio_thresh`	2
`w_mode`	both

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

Headline scores used for checkpoint selection:

Metric	Train (after unlearning)	Test (after unlearning)
Efficacy	0.566	0.642
Specificity	0.676	0.646
Harmonic mean	0.616	0.644
Relearning QA (MC)	—	0.6

Full Evaluation (baseline → unlearned)

From evaluation/score_comparison.csv:

Metric	Baseline (train)	After unlearn (train)	Baseline (test)	After unlearn (test)
QA accuracy	0.78	0.48	0.78	0.44
QA fraction	1	0.434	1	0.358
SimDom accuracy	0.82	0.56	0.9	0.64
SimDom fraction	1	0.544	1	0.6
MMLU accuracy	0.62	0.58	0.65	0.53
MMLU fraction	1	0.892	1	0.7

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