49vc631b6 commited on
Commit
c119e8a
·
verified ·
1 Parent(s): 8ebadf0

Upload folder using huggingface_hub

Browse files
Files changed (2) hide show
  1. abliteration_metadata.json +5 -5
  2. model.safetensors +3 -0
abliteration_metadata.json CHANGED
@@ -3,11 +3,11 @@
3
  "technique": "refusal_direction_ablation",
4
  "method": "advanced",
5
  "method_config": {
6
- "n_directions": 1,
7
  "direction_method": "svd",
8
  "norm_preserve": true,
9
- "regularization": 0.1,
10
- "refinement_passes": 1,
11
  "project_biases": true,
12
  "use_chat_template": true,
13
  "use_whitened_svd": false,
@@ -39,10 +39,10 @@
39
  "n_harmful_prompts": 512,
40
  "n_harmless_prompts": 512,
41
  "quality_metrics": {
42
- "perplexity": 6.109646976837031,
43
  "coherence": 1.0,
44
  "refusal_rate": 0.0,
45
- "kl_divergence": 0.018014386296272278,
46
  "spectral_certification": "RED"
47
  },
48
  "kl_contributions": {},
 
3
  "technique": "refusal_direction_ablation",
4
  "method": "advanced",
5
  "method_config": {
6
+ "n_directions": 4,
7
  "direction_method": "svd",
8
  "norm_preserve": true,
9
+ "regularization": 0.3,
10
+ "refinement_passes": 2,
11
  "project_biases": true,
12
  "use_chat_template": true,
13
  "use_whitened_svd": false,
 
39
  "n_harmful_prompts": 512,
40
  "n_harmless_prompts": 512,
41
  "quality_metrics": {
42
+ "perplexity": 6.246960088839762,
43
  "coherence": 1.0,
44
  "refusal_rate": 0.0,
45
+ "kl_divergence": 0.5171049237251282,
46
  "spectral_certification": "RED"
47
  },
48
  "kl_contributions": {},
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45be9f7aac2682e24e636d3cf7e433c5bb43e78bc4da79f5ff7ea67c56d198d0
3
+ size 807426286