Hugging Face's logo

Spaces:
d3LLM
/
dLLM_Leaderboard
Running

App Files Community
dLLM_Leaderboard
File size: 2,623 Bytes 
d473371
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
133
134
135
136
137
 
# AUP Benchmark Data
# Format: task -> method -> list of (rho, accuracy) pairs
# rho: parallelism (tokens per forward)
# accuracy: model accuracy (0-1 scale)

# Model metadata: type (AR/dLLM), foundation model, link
_meta:
  Qwen-2.5-7B-it:
    type: AR
    foundation: Qwen-2.5-7B-it
    link: https://huggingface.co/Qwen/Qwen2.5-7B-Instruct
  EAGLE-3:
    type: AR
    foundation: Llama-3.1-8B-it
    link: https://github.com/SafeAILab/EAGLE
  Dream:
    type: dLLM
    foundation: Dream-v0-it-7B
    link: https://github.com/DreamLM/Dream
  Fast-dLLM-Dream:
    type: dLLM
    foundation: Dream-v0-it-7B
    link: https://github.com/NVlabs/Fast-dLLM
  Fast-dLLM-v2:
    type: dLLM
    foundation: Qwen-2.5-7B-it
    link: https://github.com/NVlabs/Fast-dLLM/tree/main/v2
  dParallel-Dream:
    type: dLLM
    foundation: Dream-v0-it-7B
    link: https://github.com/czg1225/dParallel
  d3LLM-Dream:
    type: dLLM
    foundation: Dream-v0-it-7B
    link: https://github.com/hao-ai-lab/d3llm

GSM8K-CoT:
  Qwen-2.5-7B-it:
  - [1.0, 74.1]
  EAGLE-3:
  - [1.0, 76.57]
  - [5.12, 76.57]
  Dream:
  - [1.0, 83.94]
  Fast-dLLM-Dream:
  - [1.0, 83.68]
  - [1.44, 79.0]
  Fast-dLLM-v2:
  - [1.0, 82.82]
  - [2.21, 81.48]
  dParallel-Dream:
  - [1.0, 83.8]
  - [3.02, 82.12]
  d3LLM-Dream:
  - [1.0, 83.47]
  - [4.94, 81.36]
MATH:
  Qwen-2.5-7B-it:
  - [1.0, 41.15]
  EAGLE-3:
  - [1.0, 39.80]
  - [5.72, 39.80]
  Dream:
  - [1.0, 39.63]
  Fast-dLLM-Dream:
  - [1.0, 39.53]
  - [1.78, 38.3]
  Fast-dLLM-v2:
  - [1.0, 49.92]
  - [2.61, 48.74]
  dParallel-Dream:
  - [1.0, 39.06]
  - [2.94, 38.72]
  d3LLM-Dream:
  - [1.0, 39.38]
  - [3.92, 38.21]
MBPP-Instruct:
  Qwen-2.5-7B-it:
  - [1.0, 63.8]
  EAGLE-3:
  - [1.0, 60.20]
  - [5.69, 60.20]
  Dream:
  - [1.0, 57.2]
  Fast-dLLM-Dream:
  - [1.0, 56.38]
  - [1.2, 53.2]
  Fast-dLLM-v2:
  - [1.0, 61.23]
  - [2.04, 59.12]
  dParallel-Dream:
  - [1.0, 57.8]
  - [2.24, 55.4]
  d3LLM-Dream:
  - [1.0, 58.8]
  - [2.96, 55.60]
HumanEval-Instruct:
  Qwen-2.5-7B-it:
  - [1.0, 72.25]
  EAGLE-3:
  - [1.0, 67.58]
  - [5.98, 67.58]
  Dream:
  - [1.0, 55.2]
  Fast-dLLM-Dream:
  - [1.0, 54.86]
  - [1.33, 54.27]
  Fast-dLLM-v2:
  - [1.0, 63.2]
  - [2.58, 61.7]
  dParallel-Dream:
  - [1.0, 56.08]
  - [2.57, 54.27]
  d3LLM-Dream:
  - [1.0, 58.86]
  - [3.20, 57.10]
Long-GSM8K:
  Qwen-2.5-7B-it:
  - [1.0, 82.56]
  EAGLE-3:
  - [1.0, 80.52]
  - [5.57, 80.52]
  Dream:
  - [1.0, 78.95]
  Fast-dLLM-Dream:
  - [1.0, 78.83]
  - [1.79, 76.57]
  Fast-dLLM-v2:
  - [1.0, 82.34]
  - [2.58, 80.97]
  dParallel-Dream:
  - [1.0, 81.27]
  - [3.49, 78.56]
  d3LLM-Dream:
  - [1.0, 81.2]
  - [4.80, 77.18]