-
Notifications
You must be signed in to change notification settings - Fork 4
/
json2md.py
128 lines (119 loc) · 5.2 KB
/
json2md.py
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
json_file = "data.json"
md_file = "README.md"
import json
prefix_str = (
"# Awesome-MLLM-Safety [](https://github.com/isXinLiu/Awesome-MLLM-Safety)\n"
"A **continual** collection of papers related to safety of Multimodal Large Language Models (MLLMs).\n\n"
"We follow the definition of safety from the paper <b><q>Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructions</q></b>: <blockquote>Safety is defined as stopping models from following malicious instructions and <b>generating toxic content</b>.</blockquote>\n\n"
"<details>\n"
" <summary>The scope of our collection.</summary>\n"
" <ul>\n"
" <li>\n"
" Robustness-related wrong prediction and downstream applications (e.g., robotic/medical/legal/financial domains, anomalies detection, fake news detection) are not involved.\n"
" </li>\n"
" <li>\n"
" We care about the safety of <b>MLLMs</b>, excluding other models like text-to-image models.\n"
" </li>\n"
" <li>\n"
" We mainly focus on <b>images and text</b>, and few about other modalities like audio and videos.\n"
" </li>\n"
" </ul>\n"
"</details>\n\n\n"
"> If you find some important work missed, it would be super helpful to let me know (`[email protected]`). Thanks!\n\n"
"> If you find our survey useful for your research, please consider citing:\n\n"
"```\n"
"@article{liu:arxiv2024,\n"
" title={Safety of Multimodal Large Language Models on Images and Text},\n"
" author={Liu, Xin and Zhu, Yichen and Lan, Yunshi and Yang, Chao and Qiao, Yu},\n"
" journal={arXiv preprint arXiv:2402.00357},\n"
" year={2024}\n"
"}\n"
"```\n\n"
"Common terminologies related to safety:\n"
"<img src='./assets/terminology.jpeg' width='100%'>\n"
"Taxonomy----safety of MLLMs on images and text:\n"
"<img src='./assets/taxonomy.jpg' width='100%'>\n\n"
"**Table of Contents**\n"
"- [Evaluation](#Evaluation)\n"
"- [Attack](#Attack)\n"
"- [Defense](#Defense)\n"
"- [Other](#Other)\n"
"---\n\n"
)
paper_str_template = (
"* **{paper_title}**\n"
" * {paper_author}\n"
" * {paper_affiliation}\n"
" * [{paper_date}] {paper_link}\n"
" * {paper_notes}\n"
)
paper_str_template_code = (
"* **{paper_title}** | [Github]({paper_code_link}) \n"
" * {paper_author}\n"
" * {paper_affiliation}\n"
" * [{paper_date}] {paper_link}\n"
" * {paper_notes}\n"
)
def get_paper_str(paper):
if paper['Code'] == "":
paper_str = paper_str_template.format(
paper_date=paper['Date'],
paper_notes=paper['Notes'],
paper_link=paper['Link'],
paper_title=paper['Title'],
paper_author=paper['Author'],
paper_affiliation=paper['Affiliation'],
)
else:
paper_str = paper_str_template_code.format(
paper_date=paper['Date'],
paper_notes=paper['Notes'],
paper_link=paper['Link'],
paper_title=paper['Title'],
paper_author=paper['Author'],
paper_affiliation=paper['Affiliation'],
paper_code_link=paper['Code'],
paper_code_keyphrase=paper['Code'].split('github.com/')[1]
)
return paper_str
if __name__ == "__main__":
with open(json_file, "r") as jf:
json_data = json.load(jf)
with open(md_file,"w+") as mf:
pass
with open(md_file,"a+") as mf:
mf.write(prefix_str)
mf.write("## Evaluation\n")
for paper in json_data['Main']:
if 'benchmark' not in paper['Notes'].lower() and \
'finding' not in paper['Notes'].lower() and \
'analysis' not in paper['Notes'].lower() and \
'insight' not in paper['Notes'].lower():
continue
paper_str = get_paper_str(paper)
mf.write(paper_str)
mf.write("## Attack\n")
for paper in json_data['Main']:
if 'attack' not in paper['Notes'].lower():
continue
paper_str = get_paper_str(paper)
mf.write(paper_str)
mf.write("## Defense\n")
for paper in json_data['Main']:
if 'defense' not in paper['Notes'].lower():
continue
paper_str = get_paper_str(paper)
mf.write(paper_str)
mf.write("## Other\n")
for paper in json_data['Main']:
if 'benchmark' in paper['Notes'].lower() or \
'finding' in paper['Notes'].lower() or \
'analysis' in paper['Notes'].lower() or \
'insight' in paper['Notes'].lower():
continue
if 'attack' in paper['Notes'].lower():
continue
if 'defense' in paper['Notes'].lower():
continue
paper_str = get_paper_str(paper)
mf.write(paper_str)