-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathAIvaluateXR.html
More file actions
289 lines (212 loc) · 8.23 KB
/
Copy pathAIvaluateXR.html
File metadata and controls
289 lines (212 loc) · 8.23 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
288
289
<!DOCTYPE HTML>
<!--
Twenty by HTML5 UP
-->
<html>
<head>
<title>AIvaluateXR - Nanovis Group</title>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1, user-scalable=no" />
<link rel="stylesheet" href="assets/css/main.css" />
<noscript>
<link rel="stylesheet" href="assets/css/noscript.css" />
</noscript>
</head>
<body class="index is-preload">
<div id="page-wrapper">
<!-- Header -->
<div id="header-container"></div>
<!-- Main -->
<article id="main">
<!-- Title -->
<header class="special container">
<h2>
<strong>
AIvaluateXR: An Evaluation Framework for On-Device AI in XR with Benchmarking Results
</strong>
</h2>
<p style="margin-top:12px;">
<em>
IEEE Transactions on Visualization and Computer Graphics (IEEE TVCG), 2026.
</em>
</p>
</header>
<!-- Content -->
<section id="pub_aivaluatexr" class="wrapper style4 container special-alt">
<!-- Main Image -->
<div class="row gtr-50">
<div class="col-12 col-12-narrower">
<div class="image fit">
<img src="./images/LoXR4Nanovis.jpg" alt="AIvaluateXR"/>
<p class="fig-text"></p>
</div>
</div>
</div>
<!-- Authors -->
<div class="row gtr-50">
<div class="col-12 col-12-narrower">
<header>
<h2>Authors</h2>
<a href="https://dawarkhanuom.github.io/">Dawar Khan</a>,
<a href="https://nanovis.kaust.edu.sa/profiles">Xinyu Liu</a>,
<a href="https://cemse.kaust.edu.sa/profiles/omar-mena-facio">Omar Mena</a>,
<a href="https://cemse.kaust.edu.sa/profiles/donggang-jia">Donggang Jia</a>,
<a href="https://cemse.kaust.edu.sa/profiles/alexandre-kouyoumdjian">Alexandre Kouyoumdjian</a>,
<a href="https://cemse.kaust.edu.sa/profiles/ivan-viola">Ivan Viola</a>
</header>
<h2>Description</h2>
<p>
The deployment of large language models (LLMs) on extended reality (XR) devices has great potential to advance human–AI interaction. However, selecting the appropriate model and device for specific tasks remains challenging when performing direct on-device inference.
</p>
<p>
In this work, we present <strong>AIvaluateXR</strong>, a comprehensive evaluation framework for benchmarking LLMs running on XR devices. To demonstrate the framework, we deploy 17 selected LLMs across four XR platforms—Magic Leap 2, Meta Quest 3, Vivo X100s Pro, and Apple Vision Pro—and conduct an extensive evaluation.
</p>
<p>
Our experimental setup measures four key metrics: performance consistency, processing speed, memory usage, and battery consumption. For each of the 68 model–device pairs, we assess performance under varying string lengths, batch sizes, and thread counts, analyzing the tradeoffs for real-time XR applications.
</p>
<p>
We further propose a unified evaluation method based on 3D Pareto Optimality to identify optimal device–model combinations considering both quality and speed objectives. Additionally, we compare the efficiency of on-device LLMs with client–server and cloud-based setups and evaluate their performance on interactive XR tasks.
</p>
<p>
We believe our findings provide valuable insights for future optimization of on-device AI systems for XR. The proposed evaluation framework serves as foundational groundwork for future research and development in this emerging area.
</p>
<!-- Sources -->
<h2>Sources</h2>
<ul>
<li>
<a href="https://github.com/nanovis/AIvaluateXR" target="_blank">
Source Code / Scripts
</a>
</li>
<li>
<a href="https://youtu.be/agnLlrpLZ1o" target="_blank">
Video Demo
</a>
</li>
<li>
<a href="https://ieeexplore.ieee.org/abstract/document/10973004" target="_blank">
Poster at IEEE VR 2025
</a>
</li>
<li>
<a href="https://arxiv.org/abs/2502.15761" target="_blank">
Paper (Accepted Manuscript / arXiv)
</a>
</li>
<li>
<a href="download/supplementary_AIvaluateXR.pdf" target="_blank">
Supplementary Materials (PDF)
</a>
</li>
</ul>
<!-- Publications -->
<footer>
<h2>Publications</h2>
<p>
Dawar Khan, Xinyu Liu, Omar Mena, Donggang Jia, Alexandre Kouyoumdjian, Ivan Viola.<br />
<a href="https://arxiv.org/abs/2502.15761" target="_blank">
<strong>
AIvaluateXR: An Evaluation Framework for On-Device AI in XR with Benchmarking Results
</strong>
</a><br />
<em>
IEEE Transactions on Visualization and Computer Graphics (IEEE TVCG), 2026.
</em><br />
<span onclick="
document.getElementById('bibtex-aivaluatexr').style.display =
(document.getElementById('bibtex-aivaluatexr').style.display === 'none')
? 'block' : 'none';"
style="cursor:pointer; color:#007bff;">
<strong>[BibTeX]</strong>
</span>
</p>
<pre id="bibtex-aivaluatexr"
style="
display:none;
background-color:#f5f5f5;
padding:10px;
border:1px solid #ddd;
overflow-x:auto;">
@article{khan2026aivaluatexr,
title = {{AIvaluateXR}: An Evaluation Framework for On-Device {AI} in {XR} with Benchmarking Results},
author = {Khan, Dawar and Liu, Xinyu and Mena, Omar and Jia, Donggang and Kouyoumdjian, Alexandre and Viola, Ivan},
journal = {IEEE Transactions on Visualization and Computer Graphics (TVCG)},
year = {2026},
note = {Accepted},
url = {https://arxiv.org/abs/2502.15761}
}
</pre>
<!-- LoXR Poster -->
<p>
Xinyu Liu, Dawar Khan, Omar Mena, Donggang Jia, Alexandre Kouyoumdjian, Ivan Viola.<br />
<a href="https://doi.org/10.1109/VRW66409.2025.00252" target="_blank">
<strong>
LLMs on XR (LoXR): Performance Evaluation of LLMs Executed Locally on Extended Reality Devices
</strong>
</a><br />
<em>
IEEE VRW 2025, pp. 1212–1213.
</em><br />
<span onclick="
document.getElementById('bibtex-loxr').style.display =
(document.getElementById('bibtex-loxr').style.display === 'none')
? 'block' : 'none';"
style="cursor:pointer; color:#007bff;">
<strong>[BibTeX]</strong>
</span>
</p>
<pre id="bibtex-loxr"
style="
display:none;
background-color:#f5f5f5;
padding:10px;
border:1px solid #ddd;
overflow-x:auto;">
@inproceedings{liu2025loxr,
title = {{LLMs on XR (LoXR)}: Performance Evaluation of {LLMs} Executed Locally on Extended Reality Devices},
author = {Liu, Xinyu and Khan, Dawar and Mena, Omar and Jia, Donggang and Kouyoumdjian, Alexandre and Viola, Ivan},
booktitle = {2025 IEEE Conference on Virtual Reality and 3D User Interfaces Abstracts and Workshops (VRW)},
pages = {1212--1213},
year = {2025},
doi = {10.1109/VRW66409.2025.00252}
}
</pre>
</footer>
</div>
</div>
</section>
</article>
<!-- Footer -->
<footer id="footer">
<ul class="icons">
<li>
<a href="https://github.com/nanovis"
class="icon brands circle fa-github">
<span class="label">Github</span>
</a>
</li>
<li>
<a href="https://twitter.com/nanovis_group"
class="icon brands circle fa-twitter">
<span class="label">Twitter</span>
</a>
</li>
</ul>
<ul class="copyright">
<li>© Nanovis</li>
<li>Design: <a href="http://html5up.net">HTML5 UP</a></li>
</ul>
</footer>
</div>
<!-- Scripts -->
<script src="assets/js/jquery.min.js"></script>
<script src="assets/js/jquery.dropotron.min.js"></script>
<script src="assets/js/jquery.scrolly.min.js"></script>
<script src="assets/js/jquery.scrollex.min.js"></script>
<script src="assets/js/browser.min.js"></script>
<script src="assets/js/breakpoints.min.js"></script>
<script src="assets/js/util.js"></script>
<script src="assets/js/main.js"></script>
<script src="assets/js/load_header.js"></script>
</body>
</html>