Hugging Face's logo

Datasets:
neurips26
/
MultimodalUnlearningEvalBenchmark

Tasks:
Text Generation
Image-Text-to-Text
Modalities:
Tabular
Text
Languages:
English
Size:
< 1K
Tags:
machine-unlearning
multimodal
benchmark
evaluation
privacy
llm
License:
Dataset card Data Studio Files
xet
 Community
1
neurips26 commited on
Commit
7e36f29
·
verified ·
1 Parent(s): 531de9a

Delete analysis_results.json

Browse files
Files changed (1) hide show
  1. analysis_results.json +0 -167
analysis_results.json DELETED
@@ -1,167 +0,0 @@
1
- {
2
- "finding1_tau_matrix": {
3
- "tau_matrix": [
4
- [
5
- 1.0,
6
- 0.8518518518518519,
7
- 0.5555555555555557,
8
- -0.2592592592592593,
9
- -0.07407407407407408
10
- ],
11
- [
12
- 0.8518518518518519,
13
- 1.0,
14
- 0.5555555555555556,
15
- -0.11111111111111113,
16
- 0.0740740740740741
17
- ],
18
- [
19
- 0.5555555555555557,
20
- 0.5555555555555556,
21
- 1.0,
22
- -0.18518518518518526,
23
- -0.07407407407407407
24
- ],
25
- [
26
- -0.2592592592592593,
27
- -0.11111111111111113,
28
- -0.18518518518518526,
29
- 1.0,
30
- 0.44444444444444453
31
- ],
32
- [
33
- -0.07407407407407408,
34
- 0.0740740740740741,
35
- -0.07407407407407407,
36
- 0.44444444444444453,
37
- 1.0
38
- ]
39
- ],
40
- "p_matrix": [
41
- [
42
- 0.0,
43
- 5.6708477461516386e-06,
44
- 0.007669728020954256,
45
- 0.04311073418146568,
46
- 0.5587899692492142
47
- ],
48
- [
49
- 5.6708477461516386e-06,
50
- 0.0,
51
- 0.005390882019515592,
52
- 0.47136168054729377,
53
- 0.6810571608718818
54
- ],
55
- [
56
- 0.007669728020954256,
57
- 0.005390882019515592,
58
- 0.0,
59
- 0.384191947993784,
60
- 0.6952876854012655
61
- ],
62
- [
63
- 0.04311073418146568,
64
- 0.47136168054729377,
65
- 0.384191947993784,
66
- 0.0,
67
- 4.366826031328028e-05
68
- ],
69
- [
70
- 0.5587899692492142,
71
- 0.6810571608718818,
72
- 0.6952876854012655,
73
- 4.366826031328028e-05,
74
- 0.0
75
- ]
76
- ],
77
- "metric_names": [
78
- "FA",
79
- "RA",
80
- "MIA",
81
- "AD",
82
- "JS"
83
- ]
84
- },
85
- "finding1_contradiction": [
86
- {
87
- "method": "gradient_ascent",
88
- "FA_score": 0.6222,
89
- "RA_score": 0.62,
90
- "MIA_score": 0.2733,
91
- "AD_score": 105.0456,
92
- "JS_score": 0.0867,
93
- "FA_rank": 3,
94
- "RA_rank": 2,
95
- "MIA_rank": 1,
96
- "AD_rank": 4,
97
- "JS_rank": 3
98
- },
99
- {
100
- "method": "random_labels",
101
- "FA_score": 0.6111,
102
- "RA_score": 0.6267,
103
- "MIA_score": 0.3378,
104
- "AD_score": 41.401,
105
- "JS_score": 0.0382,
106
- "FA_rank": 1,
107
- "RA_rank": 1,
108
- "MIA_rank": 3,
109
- "AD_rank": 1,
110
- "JS_rank": 1
111
- },
112
- {
113
- "method": "finetune_retain",
114
- "FA_score": 0.6185,
115
- "RA_score": 0.58,
116
- "MIA_score": 0.3459,
117
- "AD_score": 87.1024,
118
- "JS_score": 0.1574,
119
- "FA_rank": 2,
120
- "RA_rank": 4,
121
- "MIA_rank": 4,
122
- "AD_rank": 3,
123
- "JS_rank": 4
124
- },
125
- {
126
- "method": "salun",
127
- "FA_score": 0.6222,
128
- "RA_score": 0.6,
129
- "MIA_score": 0.323,
130
- "AD_score": 51.1094,
131
- "JS_score": 0.0491,
132
- "FA_rank": 4,
133
- "RA_rank": 3,
134
- "MIA_rank": 2,
135
- "AD_rank": 2,
136
- "JS_rank": 2
137
- }
138
- ],
139
- "finding2_modality_compare": {
140
- "multimodal_mean_tau": 0.08610242435087354,
141
- "unimodal_mean_tau": 0.15800946427223814,
142
- "delta": 0.0719070399213646,
143
- "finding": "Metric agreement is 0.072 lower in multimodal settings. The two-pathway image+text architecture amplifies metric contradiction."
144
- },
145
- "finding3_reliability": {
146
- "FA": {
147
- "spearman_r": -0.4177916220015719,
148
- "p_value": 0.011230773315023548
149
- },
150
- "RA": {
151
- "spearman_r": 0.48365316307562467,
152
- "p_value": 0.0028042215785169288
153
- },
154
- "MIA": {
155
- "spearman_r": 0.22401597757064645,
156
- "p_value": 0.18903953402588844
157
- },
158
- "AD": {
159
- "spearman_r": -0.21532917220726358,
160
- "p_value": 0.20722367190215776
161
- },
162
- "JS": {
163
- "spearman_r": -0.051480051480051484,
164
- "p_value": 0.7655679786158989
165
- }
166
- }
167
- }