CoolFace
Modelpublic

KaraKaraWitch/ALIA-40b-instruct-2601-ara-heretic

sourceHugging Faceapache-2.0updated 6mo agoView on Hugging Face
0likes15downloads
Model Card

This is a decensored version of a model, made using Heretic v1.2.0 with the Arbitrary-Rank Ablation (ARA) method

Abliteration parameters

ParameterValue
start_layer_index9
end_layer_index46
preserve_good_behavior_weight0.4536
steer_bad_behavior_weight0.6028
overcorrect_relative_weight0.3384
neighbor_count7

Performance

MetricThis modelOriginal model (a model)
PIQA acc_norm0.8357Unknown
Refusals4/100100/100