Hugging Face's logo

mlx-community
/
dots.mocr-bf16

Image-Text-to-Text
MLX
Safetensors
Transformers
English
Chinese
multilingual
dots_ocr
text-generation
image-to-text
ocr
document-parse
layout
table
formula
custom_code
conversational
Eval Results
Model card Files
xet
 Community
1

Add olmOCR-bench evaluation results

#1
by pcuenq HF Staff - opened
base: refs/heads/main
←
from: refs/pr/1
Discussion Files changed
+50
-42
Files changed (1) hide show
  1. .eval_results/olmocrbench.yaml +50 -42
.eval_results/olmocrbench.yaml CHANGED
@@ -1,72 +1,80 @@
1
  - dataset:
2
  id: allenai/olmOCR-bench
3
  task_id: overall
4
- value: 83.9
5
  source:
6
- url: https://huggingface.co/papers/2603.13032
7
- name: dots.mocr technical report
8
- user: nielsr
 
9
  - dataset:
10
  id: allenai/olmOCR-bench
11
  task_id: arxiv_math
12
- value: 85.9
13
  source:
14
- url: https://huggingface.co/papers/2603.13032
15
- name: dots.mocr technical report
16
- user: nielsr
 
17
  - dataset:
18
  id: allenai/olmOCR-bench
19
- task_id: old_scans_math
20
- value: 85.5
21
  source:
22
- url: https://huggingface.co/papers/2603.13032
23
- name: dots.mocr technical report
24
- user: nielsr
 
25
  - dataset:
26
  id: allenai/olmOCR-bench
27
- task_id: table_tests
28
- value: 90.7
29
  source:
30
- url: https://huggingface.co/papers/2603.13032
31
- name: dots.mocr technical report
32
- user: nielsr
 
33
  - dataset:
34
  id: allenai/olmOCR-bench
35
- task_id: old_scans
36
- value: 48.2
37
  source:
38
- url: https://huggingface.co/papers/2603.13032
39
- name: dots.mocr technical report
40
- user: nielsr
 
41
  - dataset:
42
  id: allenai/olmOCR-bench
43
  task_id: multi_column
44
- value: 85.3
45
  source:
46
- url: https://huggingface.co/papers/2603.13032
47
- name: dots.mocr technical report
48
- user: nielsr
 
49
  - dataset:
50
  id: allenai/olmOCR-bench
51
- task_id: long_tiny_text
52
- value: 81.6
53
  source:
54
- url: https://huggingface.co/papers/2603.13032
55
- name: dots.mocr technical report
56
- user: nielsr
 
57
  - dataset:
58
  id: allenai/olmOCR-bench
59
- task_id: headers_footers
60
- value: 94.0
61
  source:
62
- url: https://huggingface.co/papers/2603.13032
63
- name: dots.mocr technical report
64
- user: nielsr
 
65
  - dataset:
66
  id: allenai/olmOCR-bench
67
- task_id: baseline
68
- value: 99.7
69
  source:
70
- url: https://huggingface.co/papers/2603.13032
71
- name: dots.mocr technical report
72
- user: nielsr
 
1
  - dataset:
2
  id: allenai/olmOCR-bench
3
  task_id: overall
4
+ value: 82.9
5
  source:
6
+ url: https://huggingface.co/datasets/pcuenq/dots.mocr-mlx-evals
7
+ name: dots.mocr MLX quantization evaluation
8
+ user: pcuenq
9
+
10
  - dataset:
11
  id: allenai/olmOCR-bench
12
  task_id: arxiv_math
13
+ value: 86.6
14
  source:
15
+ url: https://huggingface.co/datasets/pcuenq/dots.mocr-mlx-evals
16
+ name: dots.mocr MLX quantization evaluation
17
+ user: pcuenq
18
+
19
  - dataset:
20
  id: allenai/olmOCR-bench
21
+ task_id: baseline
22
+ value: 99.3
23
  source:
24
+ url: https://huggingface.co/datasets/pcuenq/dots.mocr-mlx-evals
25
+ name: dots.mocr MLX quantization evaluation
26
+ user: pcuenq
27
+
28
  - dataset:
29
  id: allenai/olmOCR-bench
30
+ task_id: headers_footers
31
+ value: 91.4
32
  source:
33
+ url: https://huggingface.co/datasets/pcuenq/dots.mocr-mlx-evals
34
+ name: dots.mocr MLX quantization evaluation
35
+ user: pcuenq
36
+
37
  - dataset:
38
  id: allenai/olmOCR-bench
39
+ task_id: long_tiny_text
40
+ value: 80.8
41
  source:
42
+ url: https://huggingface.co/datasets/pcuenq/dots.mocr-mlx-evals
43
+ name: dots.mocr MLX quantization evaluation
44
+ user: pcuenq
45
+
46
  - dataset:
47
  id: allenai/olmOCR-bench
48
  task_id: multi_column
49
+ value: 84.6
50
  source:
51
+ url: https://huggingface.co/datasets/pcuenq/dots.mocr-mlx-evals
52
+ name: dots.mocr MLX quantization evaluation
53
+ user: pcuenq
54
+
55
  - dataset:
56
  id: allenai/olmOCR-bench
57
+ task_id: old_scans
58
+ value: 48.1
59
  source:
60
+ url: https://huggingface.co/datasets/pcuenq/dots.mocr-mlx-evals
61
+ name: dots.mocr MLX quantization evaluation
62
+ user: pcuenq
63
+
64
  - dataset:
65
  id: allenai/olmOCR-bench
66
+ task_id: old_scans_math
67
+ value: 83.6
68
  source:
69
+ url: https://huggingface.co/datasets/pcuenq/dots.mocr-mlx-evals
70
+ name: dots.mocr MLX quantization evaluation
71
+ user: pcuenq
72
+
73
  - dataset:
74
  id: allenai/olmOCR-bench
75
+ task_id: table_tests
76
+ value: 88.4
77
  source:
78
+ url: https://huggingface.co/datasets/pcuenq/dots.mocr-mlx-evals
79
+ name: dots.mocr MLX quantization evaluation
80
+ user: pcuenq