Hugging Face's logo

JohnP1
/
d1a-e4b-it

Text Classification
PEFT
Safetensors
English
Japanese
d1a
decision-model
calibration
lora
gemma4
typesafe
system-one
pull-requests
chat
on-device
Model card Files
xet
 Community
d1a-e4b-it
File size: 2,090 Bytes
2a87cbb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
{
 "model": "D1A-E4B-it v0.6",
 "temperature": 1.551,
 "method": "Every D1A suite (scripts/quality_gate.py --all-suites --sequential, then scripts/decide.py) on a Mac mini M4, 2026-10-17, against D1A-E2B-it v0.6, both as served: MLX 8-bit unmerged chat builds with media, each at its own temperature. Paired bootstrap 95% intervals. Live PR-labeler replay and hand-labelled PRs are private; aggregates only.",
 "suites": {
  "decision-v7": [
   1264,
   84.8,
   85.7,
   "+0.9 [-0.7,+2.5]"
  ],
  "transfer-v4": [
   656,
   75.0,
   77.4,
   "+2.4 [-0.9,+5.9]"
  ],
  "hard-v1": [
   1083,
   69.2,
   75.7,
   "+6.6 [+4.0,+9.2]"
  ],
  "devtools-v1": [
   1074,
   71.1,
   74.3,
   "+3.2 [+0.9,+5.4]"
  ],
  "documents-v1": [
   920,
   86.3,
   88.6,
   "+2.3 [+0.5,+4.2]"
  ],
  "ja-jglue:development": [
   1500,
   78.6,
   80.8,
   "+2.2 [+0.5,+4.0]"
  ],
  "ja-jglue:test": [
   1500,
   76.4,
   78.6,
   "+2.2 [+0.3,+4.1]"
  ],
  "pr-labels:test": [
   2045,
   77.4,
   79.1,
   "+1.8 [+0.5,+3.1]"
  ],
  "pr-labels:test sev+offsets": [
   953,
   64.3,
   69.5,
   "+5.1 [+3.0,+7.3]"
  ],
  "pr-labels:development": [
   2030,
   78.8,
   79.6,
   "+0.8 [-0.4,+2.1]"
  ],
  "pr-labels:test-ja": [
   193,
   74.6,
   77.7,
   "+3.1 [-1.0,+7.2]"
  ],
  "routing:factory-development": [
   640,
   92.3,
   93.6,
   "+1.2 [-0.7,+3.2]"
  ],
  "routing:generic-development": [
   270,
   97.0,
   97.4,
   "+0.4 [-1.5,+2.2]"
  ],
  "external:typesafe-v1": [
   102,
   69.6,
   86.3,
   "+16.7 [+8.8,+25.2]"
  ],
  "external:wanli-v2": [
   1002,
   68.9,
   72.7,
   "+3.8 [+1.4,+6.2]"
  ],
  "night2:dates": [
   900,
   87.0,
   90.8,
   "+3.8 [+1.7,+5.8]"
  ],
  "night2:unknowable": [
   525,
   69.5,
   66.7,
   "-2.9 [-5.5,-0.2]"
  ]
 },
 "columns": [
  "n",
  "E2B-it v0.6 accuracy %",
  "E4B-it v0.6 accuracy %",
  "delta [95% CI]"
 ],
 "pooled_delta": "+2.5 [+1.7, +3.4] over 21 suite lines and 2 live",
 "latency_ratio_vs_e2b_it": 2.94,
 "media": "14 photo/voice/video demo requests, 28 answers, all answered; 7 differ from E2B-it, 6 better and 1 worse by hand read"
}