rain1955 commited on
Commit
feb586c
·
verified ·
1 Parent(s): 8f44761

SEO: add multilingual tags, model-index metrics, thumbnail, expand search coverage

Browse files
Files changed (1) hide show
  1. README.md +33 -2
README.md CHANGED
@@ -1,25 +1,56 @@
1
  ---
2
  language:
3
  - en
 
 
 
4
  tags:
5
  - emotion-vectors
6
  - interpretability
7
  - mechanistic-interpretability
8
  - replication
9
  - gemma4
 
10
  - anthropic
11
  - valence-arousal
12
  - PCA
13
  - logit-lens
 
 
14
  - emotion
15
  - functional-emotions
16
  - AI-safety
 
 
 
 
17
  license: mit
18
  library_name: transformers
19
  pipeline_tag: text-generation
20
  base_model: google/gemma-4-E4B-it
21
- datasets:
22
- - custom
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
23
  ---
24
 
25
  # Replicating Anthropic's Emotion Vectors on an Open-Source 4B Model
 
1
  ---
2
  language:
3
  - en
4
+ - zh
5
+ - ko
6
+ - es
7
  tags:
8
  - emotion-vectors
9
  - interpretability
10
  - mechanistic-interpretability
11
  - replication
12
  - gemma4
13
+ - google
14
  - anthropic
15
  - valence-arousal
16
  - PCA
17
  - logit-lens
18
+ - linear-probe
19
+ - probing
20
  - emotion
21
  - functional-emotions
22
  - AI-safety
23
+ - neuroscience
24
+ - circumplex-model
25
+ - activation-extraction
26
+ - residual-stream
27
  license: mit
28
  library_name: transformers
29
  pipeline_tag: text-generation
30
  base_model: google/gemma-4-E4B-it
31
+ thumbnail: results/fig1_pca_scatter.png
32
+ model-index:
33
+ - name: emotion-vector-replication
34
+ results:
35
+ - task:
36
+ type: text-generation
37
+ name: Emotion Vector Extraction
38
+ metrics:
39
+ - name: PC1 Variance (Valence)
40
+ type: variance_explained
41
+ value: 0.422
42
+ - name: PC2 Variance (Arousal)
43
+ type: variance_explained
44
+ value: 0.183
45
+ - name: Total Variance (PC1+PC2)
46
+ type: variance_explained
47
+ value: 0.605
48
+ - name: Emotions Tested
49
+ type: count
50
+ value: 9
51
+ - name: Stories Generated
52
+ type: count
53
+ value: 1002
54
  ---
55
 
56
  # Replicating Anthropic's Emotion Vectors on an Open-Source 4B Model