-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
287 lines (265 loc) · 14.8 KB
/
Copy pathindex.html
File metadata and controls
287 lines (265 loc) · 14.8 KB
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
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<meta name="description"
content="Benchmarking Trustworthiness of Multimodal Large Language Models">
<meta name="keywords" content="MultiTrust, MMTrustEval">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title> MultiTrust </title>
<!-- Global site tag (gtag.js) - Google Analytics -->
<!-- <script async src="https://www.googletagmanager.com/gtag/js?id=G-PYVRSFMDRL"></script> -->
<!-- <script>
window.dataLayer = window.dataLayer || [];
function gtag() {
dataLayer.push(arguments);
}
/Users/panlu/Library/Mobile Documents/com~apple~CloudDocs/ImageMath/visual-mathqa-server/data_final/images
gtag('js', new Date());
gtag('config', 'G-PYVRSFMDRL');
</script> -->
<link rel="icon" href="./static/images/multitrust.png">
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
<link rel="stylesheet" href="./static/css/bulma.min.css">
<link rel="stylesheet" href="./static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="./static/css/bulma-slider.min.css">
<link rel="stylesheet" href="./static/css/fontawesome.all.min.css">
<link rel="stylesheet" href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="./static/css/index.css">
<link rel="stylesheet" href="./static/css/leaderboard.css">
<!-- <link href="https://unpkg.com/tabulator-tables@5.5.2/dist/css/tabulator_bulma.min.css" rel="stylesheet">
<script type="text/javascript" src="https://unpkg.com/tabulator-tables@5.5.2/dist/js/tabulator.min.js"></script> -->
<script type="text/javascript" src="static/js/sort-table.js" defer></script>
<script defer src="./static/js/fontawesome.all.min.js"></script>
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script src="./static/js/bulma-carousel.min.js"></script>
<script src="./static/js/bulma-slider.min.js"></script>
<script src="./data/model_scores.js" defer></script>
<script src="./static/js/leaderboard_multitrust.js"></script>
</head>
<body>
<nav class="navbar" role="navigation" aria-label="main navigation">
<div class="navbar-brand">
<a role="button" class="navbar-burger" aria-label="menu" aria-expanded="false">
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
</a>
</div>
<div class="navbar-menu">
<div class="navbar-start" style="flex-grow: 1; justify-content: center;">
<!-- <a class="navbar-item" href="https://keunhong.com">
<span class="icon">
<i class="fas fa-home"></i>
</span>
</a> -->
<!-- @PAN TODO: consider adding links? -->
<div class="navbar-item has-dropdown is-hoverable">
<a class="navbar-link">
More Research
</a>
<div class="navbar-dropdown">
<!-- <a class="navbar-item" href="https://chameleon-llm.github.io/">
<b>Chameleon</b> <p style="font-size:18px; display: inline; margin-left: 5px;">🔥</p>
</a> -->
</div>
</div>
</div>
</div>
</nav>
<div class="header">
<div class="container is-widescreen">
<div class="columns is-centered">
<div class="column has-text-centered">
<h1>
<img src="static/images/multitrust.png" style="width:1.8em;vertical-align: middle" alt="Logo"/>
<span style="vertical-align: middle"> MultiTrust </span>
</h1>
<h2 class="is-size-3">A Comprehensive Benchmark Towards Trustworthy Multimodal Large Language Models</h2>
<div class="is-size-5 publication-authors">
<br>
<span class="author-block">
<a href="https://zycheiheihei.github.io/">Yichi Zhang</a><sup style="color:#1db2ec">1</sup>,</span>
<span class="author-block">
<a href="https://multi-trust.github.io/">Yao Huang</a><sup style="color:#f31a1a">2</sup>,</span>
<span class="author-block">
<a href="https://multi-trust.github.io/">Yitong Sun</a><sup style="color:#f31a1a">2</sup>,</span>
<span class="author-block">
<a href="https://multi-trust.github.io/">Chang Liu</a><sup style="color:#ffac33">3</sup>,</span>
<span class="author-block">
<a href="https://persistz.github.io/">Zhe Zhao</a><sup style="color:#ec1de2">4</sup>,</span>
<span class="author-block">
<a href="https://multi-trust.github.io/">Zhengwei Fang</a><sup style="color:#1db2ec">1</sup>,</span>
<span class="author-block">
<a href="https://multi-trust.github.io/">Yifan Wang</a><sup style="color:#1db2ec">1</sup>,</span>
<span class="author-block">
<a href="https://huanranchen.github.io/">Huanran Chen</a><sup style="color:#1db2ec">1</sup>,</span>
<span class="author-block">
<a href="https://ml.cs.tsinghua.edu.cn/~xiaoyang/">Xiao Yang</a><sup style="color:#1db2ec">1</sup>,</span>
<span class="author-block">
<a href="https://sites.google.com/site/xingxingwei1988/?pli=1">Xingxing Wei</a><sup style="color:#f31a1a">2</sup>,</span>
<span class="author-block">
<a href="https://www.suhangss.me/">Hang Su</a><sup style="color:#1db2ec">1</sup>,</span>
<span class="author-block">
<a href="https://ml.cs.tsinghua.edu.cn/~yinpeng/">Yinpeng Dong</a><sup style="color:#1db2ec">1</sup>,</span>
<span class="author-block">
<a href="https://ml.cs.tsinghua.edu.cn/~jun/index.shtml">Jun Zhu</a><sup style="color:#1db2ec">1</sup></span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block"><sup style="color:#1db2ec">1</sup>Tsinghua University,</span>
<span class="author-block"><sup style="color:#f31a1a">2</sup>Beihang University,</span>
<span class="author-block"><sup style="color:#ffac33">3</sup>Shanghai Jiaotong University,</span>
<span class="author-block"><sup style="color:#ec1de2">4</sup>RealAI</span>
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- PDF Link. -->
<span class="link-block">
<!-- @PAN TODO: change links -->
<a href="https://arxiv.org/pdf/2406.07057"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Paper</span>
</a>
</span>
<span class="link-block">
<a href="https://arxiv.org/abs/2406.07057"
class="external-link button is-normal is-rounded is-dark">
<!-- <a href="https://lupantech.github.io/papers/arxiv23_mathvista.pdf"
class="external-link button is-normal is-rounded is-dark"> -->
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>arXiv</span>
</a>
</span>
<!-- Code Link. -->
<span class="link-block">
<a href="https://github.com/thu-ml/MMTrustEval"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code</span>
</a>
</span>
<!-- Dataset Link. -->
<span class="link-block">
<a href="https://huggingface.co/datasets/thu-ml/MultiTrust"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<!-- <i class="far fa-images"></i> -->
<p style="font-size:18px">🤗</p>
<!-- 🔗 -->
</span>
<span>Dataset</span>
</a>
</span>
<!-- Leaderboard Link. -->
<span class="link-block">
<a href="https://multi-trust.github.io/#leaderboard"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<p style="font-size:18px">🏆</p>
</span>
<span>Leaderboard</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
<section class="section">
<div class="container" style="margin-bottom: 2vh;margin-top: 2vh">
<!-- Abstract. -->
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Introduction</h2>
<img src="static/images/framework.png" alt="framework" width="95%"/>
<div class="content has-text-justified">
<p>
Despite the superior capabilities of <b>Multimodal Large Language Models (MLLMs)</b> across diverse tasks, they still face significant trustworthiness challenges. Yet, current literature on the assessment of trustworthy MLLMs remains limited, lacking <b>a holistic evaluation</b> to offer thorough insights into future improvements.
</p>
<p>
In this work, we establish <b>MultiTrust</b>, the first comprehensive and unified benchmark on the trustworthiness of MLLMs across five primary aspects: <i>truthfulness</i>, <i>safety</i>, <i>robustness</i>, <i>fairness</i>, and <i>privacy</i>. Our benchmark employs a rigorous evaluation strategy that addresses both <b>multimodal risks</b> and <b>cross-modal impacts</b>, encompassing 32 diverse tasks with self-curated datasets.
</p>
<p>
Extensive experiments with 21 modern MLLMs reveal some previously unexplored trustworthiness issues and risks, highlighting the complexities introduced by the multimodality and underscoring the necessity for advanced methodologies to enhance their reliability. For instance, typical proprietary models still struggle with the perception of visually confusing images and are vulnerable to multimodal jailbreaking and adversarial attacks; MLLMs are more inclined to disclose privacy in text and reveal ideological and cultural biases even when paired with irrelevant images in inference, indicating that the multimodality amplifies the internal risks from base LLMs.
</p>
</div>
</div>
</div>
<!--/ Abstract. -->
</div>
</section>
<section class="section">
<div class="container" style="margin-bottom: 2vh;">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3" id="task">Task Pool</h2>
<!-- <img src="static/images/task-description.png" alt="task-description" width="70%"/> -->
<img src="static/images/task-list.png" alt="task-list" width="70%"/>
<div class="content has-text-justified">
<p class="mt-3">
<b>Task Overview:</b> We organize a two-level taxonomy containing 10 sub-aspects to better categorize the target behaviors to be evaluated. Based on the taxonomy, we curate 32 diverse tasks to cover realistic and comprehensive scenarios with trustworthy risks, including generative and discriminative, multimodal and text-only ones, as summarized above. To tackle the current lack of datasets dedicated for various scenarios under these sub-aspects, we construct 20 datasets based on the existing text, image, and multimodal datasets by adapting prompts, images, and annotations with both manual efforts and automatic methods. We further propose 8 novel datasets from scratch by collecting images from the Internet or synthesizing images with Stable Diffusion and other algorithms specifically for the designed tasks.<br>
<img src="static/images/icon/circle-xmark-regular.svg" style="width:1.0em;vertical-align: middle" alt="circle-xmark"/> : off-the-shelf datasets from prior work; <img src="static/images/icon/circle-plus-solid.svg" style="width:1.0em;vertical-align: middle" alt="circle-plus"/> : datasets adapted to new tasks with additional images, prompts, and annotations; <img src="static/images/icon/circle-check-solid.svg" style="width:1.0em;vertical-align: middle" alt="circle-check"/> : datasets constructed from scratch. <img src="static/images/icon/file-image-regular.svg" style="width:1.0em;vertical-align: middle" alt="file-image"/> : tasks for revealing multimodal risks; <img src="static/images/icon/file-lines-regular.svg" style="width:1.0em;vertical-align: middle" alt="file-text"/> : tasks for studying cross-modal impacts. <img src="static/images/icon/circle-regular.svg" style="width:1.0em;vertical-align: middle" alt="circle"/> : rule-based evaluation (e.g., keywords matching); <img src="static/images/icon/circle-solid.svg" style="width:1.0em;vertical-align: middle" alt="circle-solid"/> : automatic evaluation by GPT-4 or other classifiers; <img src="static/images/icon/circle-half-stroke-solid.svg" style="width:1.0em;vertical-align: middle" alt="circle-left"/> : mixture evaluation. ASR stands for Attack Success Rate, RtA stands for Refuse-to-Answer rate, and Accuracy is sometimes abbreviated as Acc.
</p>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container" style="margin-bottom: 2vh;">
<div class="columns is-centered">
<div class="column is-four-fifths has-text-centered content">
<div class="content">
<p class="title is-3" id="leaderboard">Leaderboard in MultiTrust (Updating...)
</p>
<div id="multitrust_leaderboard"></div>
</div>
</div>
</div>
</div>
</section>
<!-- @PAN TODO: bibtex -->
<section class="section" id="BibTeX">
<div class="container is-max-desktop content" style="margin-bottom: 2vh;">
<h2 class="title is-3 has-text-centered">BibTeX</h2>
<pre><code>
@misc{zhang2024benchmarking,
title={Benchmarking Trustworthiness of Multimodal Large Language Models: A Comprehensive Study},
author={Yichi Zhang and Yao Huang and Yitong Sun and Chang Liu and Zhe Zhao and Zhengwei Fang and
Yifan Wang and Huanran Chen and Xiao Yang and Xingxing Wei and Hang Su and Yinpeng Dong and
Jun Zhu},
year={2024},
eprint={2406.07057},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
</code></pre>
</div>
</section>
<footer class="footer">
<!-- <div class="container"> -->
<div class="content has-text-centered">
</div>
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p>
This website is website adapted from <a href="https://mathvista.github.io/"><span class="mathvista">MathVista</span></a>, licensed under a <a rel="license"
href="http://creativecommons.org/licenses/by-sa/4.0/">Creative
Commons Attribution-ShareAlike 4.0 International License</a>.
</p>
</div>
</div>
</div>
<!-- </div> -->
</footer>
</body>
</html>