Datasets:

Modalities:
Image
Text
Formats:
parquet
Languages:
Korean
Size:
< 1K
ArXiv:
Libraries:
Datasets
pandas
License:
kimyoungjune commited on
Commit
3ad6ff0
โ€ข
1 Parent(s): 2a44024

Upload 2 files

Browse files
Files changed (3) hide show
  1. .gitattributes +1 -0
  2. K-DTCBench.csv +3 -0
  3. K-DTCBench.md +95 -0
.gitattributes CHANGED
@@ -57,3 +57,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
57
  # Video files - compressed
58
  *.mp4 filter=lfs diff=lfs merge=lfs -text
59
  *.webm filter=lfs diff=lfs merge=lfs -text
 
 
57
  # Video files - compressed
58
  *.mp4 filter=lfs diff=lfs merge=lfs -text
59
  *.webm filter=lfs diff=lfs merge=lfs -text
60
+ K-DTCBench.csv filter=lfs diff=lfs merge=lfs -text
K-DTCBench.csv ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3ee9d974c425105a308d0bbdaa3ee05fe82db372ef69e9185b249deec6f48da1
3
+ size 12886645
K-DTCBench.md ADDED
@@ -0,0 +1,95 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-nc-4.0
3
+ ---
4
+
5
+ # K-DTCBench
6
+
7
+ We introduce **K-DTCBench**, a newly developed Korean benchmark featuring both computer-generated and handwritten documents, tables, and charts.
8
+ It consists of 80 questions for each image type and two questions per image, summing up to 240 questions in total.
9
+ This benchmark is designed to evaluate whether vision-language models can process images in different formats and be applicable for diverse domains.
10
+ All images are generated with made-up values and statements for evaluation purposes only. We scanned hand-written documents/tables/charts, or created digital objects with matplotlib library to build K-DTCBench.
11
+ The proportions of digital and hand-written images are equal, each constituting 50%.
12
+
13
+
14
+ For more details, Please refer to the [VARCO-VISION technical report(Coming Soon)]().
15
+
16
+ <table>
17
+ <tr>
18
+ <th>Category</th>
19
+ <th>Image</th>
20
+ <th>K-DTCBench</th>
21
+ </tr>
22
+ <tr>
23
+ <td align="center">document</td>
24
+ <td width=350><img src="https://cdn-uploads.huggingface.co/production/uploads/624ceaa38746b2f5773c2d1c/Ipi4HR73P-PDC5XcgP3WF.png"></td>
25
+ <td>
26
+ <strong>question:</strong> ๋ณด๊ณ ์„œ์˜ ์ฃผ์š” ๋‚ด์šฉ์ด ์•„๋‹Œ ๊ฒƒ์€ ๋ฌด์—‡์ธ๊ฐ€์š”?
27
+ <br>
28
+ <strong>A:</strong> ์•ˆ์ „ ์ธํ”„๋ผ ํ™•์ถฉ
29
+ <br>
30
+ <strong>B:</strong> ์žฌ๋‚œ ๋ฐ ์‚ฌ๊ณ  ์˜ˆ๋ฐฉ ์ฒด๊ณ„ ๊ตฌ์ถ•
31
+ <br>
32
+ <strong>C:</strong> ์‹œ๋ฏผ ์•ˆ์ „ ๊ต์œก ๊ฐ•ํ™”
33
+ <br>
34
+ <strong>D:</strong> ๊ธด๊ธ‰ ๋Œ€์‘ ์‹œ์Šคํ…œ ๊ฐœ์„ 
35
+ </td>
36
+ </tr>
37
+ <tr>
38
+ <td align="center">table</td>
39
+ <td width=350><img src="https://cdn-uploads.huggingface.co/production/uploads/624ceaa38746b2f5773c2d1c/dz_FuPnpZ5P4P3LEB5PZ0.png"></td>
40
+ <td>
41
+ <strong>question:</strong> ์ธํ”„๋ผ ๊ตฌ์ถ• ํ•ญ๋ชฉ์˜ ์ ์ˆ˜๋Š” ๋ช‡ ์ ์ธ๊ฐ€์š”?
42
+ <br>
43
+ <strong>A:</strong> 4
44
+ <br>
45
+ <strong>B:</strong> 6
46
+ <br>
47
+ <strong>C:</strong> 8
48
+ <br>
49
+ <strong>D:</strong> 10
50
+ </td>
51
+ </tr>
52
+ <tr>
53
+ <td align="center">chart</td>
54
+ <td width=350><img src="https://cdn-uploads.huggingface.co/production/uploads/624ceaa38746b2f5773c2d1c/IbNMPPgd974SbCAsz6zIS.png"></td>
55
+ <td>
56
+ <strong>question:</strong> ์ง์žฅ์ธ๋“ค์ด ํ‡ด๊ทผ ํ›„ ๋‘ ๋ฒˆ์งธ๋กœ ์„ ํ˜ธํ•˜๋Š” ํ™œ๋™์€ ๋ฌด์—‡์ธ๊ฐ€์š”?
57
+ <br>
58
+ <strong>A:</strong> ์šด๋™
59
+ <br>
60
+ <strong>B:</strong> ์—ฌ๊ฐ€ํ™œ๋™
61
+ <br>
62
+ <strong>C:</strong> ์ž๊ธฐ๊ฐœ๋ฐœ
63
+ <br>
64
+ <strong>D:</strong> ํœด์‹
65
+ </td>
66
+ </tr>
67
+ </table>
68
+
69
+ <br>
70
+
71
+ ## Inference Prompt
72
+ ```
73
+ <image>
74
+ {question}
75
+ Options: A: {A}, B: {B}, C: {C}, D: {D}
76
+
77
+ ์ฃผ์–ด์ง„ ์„ ํƒ์ง€ ์ค‘ ํ•ด๋‹น ์˜ต์…˜์˜ ๋ฌธ์ž๋กœ ์ง์ ‘ ๋‹ตํ•˜์„ธ์š”.
78
+ ```
79
+
80
+ <br>
81
+
82
+ ## Results
83
+ Below are the evaluation results of various vision-language models, including [VARCO-VISION-14B]() on K-DTCBench.
84
+
85
+ | | VARCO-VISION-14B | Pangea-7B | Pixtral-12B | Molmo-7B-D | Qwen2-VL-7B-Instruct | LLaVA-One-Vision-7B |
86
+ | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
87
+ | K-DTCBench | **84.58** | 48.33 | 27.50 | 45.83 | 75.00 | 52.91 |
88
+
89
+ <br>
90
+
91
+ ## Citation
92
+ (bibtex will be updated soon..) If you use K-DTCBench in your research, please cite the following:
93
+ ```
94
+
95
+ ```