Hugging Face's logo

JackYoung27
/
writesae-ckpts

Feature Extraction
PyTorch
English
sparse-autoencoder
interpretability
mechanistic-interpretability
gated-deltanet
mamba
rwkv
linear-attention
state-space-model
Model card Files
xet
 Community
writesae-ckpts
File size: 3,187 Bytes
f1850af
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
{
  "config": {
    "feature_id": 412,
    "preferred_token_id": 98818,
    "preferred_token_str": "\u7a7a\u95f4",
    "a_star": 0.3105553686618805,
    "concentration": 14.4923095703125
  },
  "install": {
    "n": 600,
    "by_magnitude": {
      "0.0000": {
        "n": 150,
        "median_delta_logp_pf": 0.0,
        "mean_delta_logp_pf": 0.0,
        "ci95_median": [
          0.0,
          0.0
        ],
        "paired_wilcoxon_p": null,
        "rank_pf_native_median": 151480.0,
        "rank_pf_patched_median": 151480.0
      },
      "1.0000": {
        "n": 150,
        "median_delta_logp_pf": 3.147125244140625e-05,
        "mean_delta_logp_pf": 0.006514116923014323,
        "ci95_median": [
          -0.009778976440429688,
          0.02117633819580078
        ],
        "paired_wilcoxon_p": 0.6653975264837414,
        "rank_pf_native_median": 151480.0,
        "rank_pf_patched_median": 151617.0
      },
      "2.0000": {
        "n": 150,
        "median_delta_logp_pf": 0.008390426635742188,
        "mean_delta_logp_pf": 0.010637429555257162,
        "ci95_median": [
          -0.008724212646484375,
          0.024947643280029297
        ],
        "paired_wilcoxon_p": 0.2752582491238841,
        "rank_pf_native_median": 151480.0,
        "rank_pf_patched_median": 152283.0
      },
      "4.0000": {
        "n": 150,
        "median_delta_logp_pf": 0.015584945678710938,
        "mean_delta_logp_pf": 0.015919570922851563,
        "ci95_median": [
          -0.0032224655151367188,
          0.059861183166503906
        ],
        "paired_wilcoxon_p": 0.14514399842092737,
        "rank_pf_native_median": 151480.0,
        "rank_pf_patched_median": 153069.5
      }
    }
  },
  "erase": {
    "n": 150,
    "median_delta_logp_pf": -0.11577987670898438,
    "mean_delta_logp_pf": -0.3492596944173177,
    "ci95_median": [
      -0.26457786560058594,
      -0.04221153259277344
    ],
    "paired_wilcoxon_p": 1.0653819862459425e-06,
    "rank_pf_native_median": 68485.0,
    "rank_pf_patched_median": 77444.0
  },
  "dose_response": {
    "magnitudes": [
      0.0,
      0.25,
      0.5,
      1.0,
      2.0,
      4.0
    ],
    "n_per_magnitude": [
      50,
      50,
      50,
      50,
      50,
      50
    ],
    "median_delta": [
      0.0,
      0.0034532546997070312,
      0.004258155822753906,
      -0.0011653900146484375,
      -0.014678001403808594,
      -0.021813392639160156
    ],
    "ci_lo": [
      0.0,
      -0.01807670593261718,
      -0.011584281921386719,
      -0.023235321044921875,
      -0.043369293212890625,
      -0.049655914306640625
    ],
    "ci_hi": [
      0.0,
      0.013463020324707031,
      0.015409469604492188,
      0.021053314208984375,
      0.0078125,
      0.007886886596679688
    ],
    "linear_fit": {
      "alpha": 0.0037066050938197545,
      "beta": -0.0067335489158806185,
      "r_squared": 0.8939612739226357,
      "aic": -64.9996182592636
    },
    "saturating_fit": {
      "alpha": -186092.974094106,
      "beta": 34553208.75298938,
      "r_squared": 0.817700677085506,
      "aic": -61.74854466042849
    }
  },
  "validation": {
    "n": 0,
    "worst_logit_diff": 0.0
  }
}