-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathindex.html
More file actions
469 lines (446 loc) · 20.6 KB
/
index.html
File metadata and controls
469 lines (446 loc) · 20.6 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
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>The Road to Convergence — CVPR 2026 Tutorial</title>
<style>
* { margin: 0; padding: 0; box-sizing: border-box; }
html { scroll-behavior: smooth; }
body {
background: #fff;
color: #000;
font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Helvetica, Arial, sans-serif;
line-height: 1.75;
}
/* NAV */
nav {
position: fixed; top: 0; left: 0; right: 0; z-index: 100;
background: #fff;
border-bottom: 1px solid #ddd;
padding: 0 2rem;
}
nav .inner {
max-width: 960px; margin: 0 auto;
display: flex; align-items: center; justify-content: space-between;
height: 52px;
}
nav .logo {
font-size: 0.95rem; font-weight: 600; color: #000;
text-decoration: none;
}
nav .logo span { color: #b8322e; }
nav ul { list-style: none; display: flex; gap: 1.6rem; }
nav ul li a {
color: #000; text-decoration: none; font-size: 0.8rem;
font-weight: 500; transition: color 0.2s;
}
nav ul li a:hover { color: #b8322e; }
/* HERO */
.hero {
min-height: 92vh;
display: flex; align-items: center;
padding: 8rem 2rem 5rem;
}
.hero-content { max-width: 960px; margin: 0 auto; }
.hero-badge {
display: inline-block;
font-size: 0.72rem; font-weight: 600;
letter-spacing: 0.1em; text-transform: uppercase;
color: #b8322e;
margin-bottom: 1.5rem;
}
.hero h1 {
font-size: clamp(2rem, 4.5vw, 3rem);
line-height: 1.2; letter-spacing: -0.02em;
margin-bottom: 1.5rem; max-width: 720px; font-weight: 700;
}
.hero h1 em {
font-style: italic;
color: #b8322e;
}
.hero p.subtitle {
font-size: 1rem; color: #000; max-width: 600px;
line-height: 1.8; margin-bottom: 3rem;
}
.hero-meta {
display: flex; gap: 3rem; flex-wrap: wrap;
border-top: 1px solid #ddd; padding-top: 1.5rem;
}
.hero-meta .item { display: flex; flex-direction: column; gap: 0.15rem; }
.hero-meta .label {
font-size: 0.68rem; text-transform: uppercase; letter-spacing: 0.08em;
color: #000; font-weight: 500;
}
.hero-meta .value { font-size: 0.9rem; font-weight: 600; }
.hero-meta .value.highlight { color: #b8322e; }
/* SECTIONS */
section { padding: 5rem 2rem; }
.container { max-width: 960px; margin: 0 auto; }
.section-label {
font-size: 0.7rem; font-weight: 600;
text-transform: uppercase; letter-spacing: 0.12em;
color: #b8322e; margin-bottom: 0.6rem;
}
.section-title {
font-size: clamp(1.5rem, 3vw, 2rem);
line-height: 1.25; letter-spacing: -0.01em;
margin-bottom: 0.6rem; font-weight: 700;
}
.section-desc {
color: #000; font-size: 0.92rem; max-width: 580px;
line-height: 1.8; margin-bottom: 2.5rem;
}
/* ABOUT */
#about { border-top: 1px solid #ddd; }
.about-grid { display: grid; grid-template-columns: 1fr 1fr 1fr; gap: 2rem; }
.about-card { padding-left: 1rem; border-left: 2px solid #ddd; }
.about-card:first-child { border-left-color: #b8322e; }
.about-card h3 { font-size: 1rem; font-weight: 600; margin-bottom: 0.4rem; }
.about-card p { font-size: 0.85rem; color: #000; line-height: 1.75; }
/* SCHEDULE */
#schedule { border-top: 1px solid #ddd; border-bottom: 1px solid #ddd; }
.schedule-list { display: flex; flex-direction: column; }
.schedule-item {
display: grid; grid-template-columns: 140px 1fr; gap: 2rem;
padding: 1.5rem 0; border-bottom: 1px solid #ddd;
}
.schedule-item:last-child { border-bottom: none; }
.schedule-time { font-size: 0.75rem; color: #000; padding-top: 0.2rem; }
.schedule-body h3 { font-size: 1.05rem; font-weight: 600; margin-bottom: 0.4rem; }
.schedule-body p { font-size: 0.85rem; color: #000; line-height: 1.75; max-width: 640px; }
/* SPEAKERS */
#speakers { border-top: 1px solid #ddd; }
.speakers-grid {
display: grid; grid-template-columns: 1fr 1fr; gap: 1px;
background: #ddd; border: 1px solid #ddd;
}
.speaker-card {
background: #fff; padding: 1.8rem;
display: flex; flex-direction: column; gap: 0.4rem;
}
.speaker-card:last-child:nth-child(odd) { grid-column: 1 / -1; }
.speaker-card h3 { font-size: 1rem; font-weight: 600; }
.speaker-role { font-size: 0.8rem; color: #000; }
.speaker-tag {
font-size: 0.68rem; font-weight: 600;
text-transform: uppercase; letter-spacing: 0.06em;
color: #b8322e; margin-top: 0.2rem;
}
.speaker-bio { font-size: 0.82rem; color: #000; line-height: 1.75; margin-top: 0.3rem; }
.speaker-link {
font-size: 0.8rem; color: #b8322e; text-decoration: none; font-weight: 500;
}
.speaker-link:hover { text-decoration: underline; }
/* TOPICS */
#topics { border-top: 1px solid #ddd; border-bottom: 1px solid #ddd; }
.topics-list { display: flex; flex-direction: column; }
.topic-item {
display: grid; grid-template-columns: 2rem 1fr; gap: 1rem;
padding: 1.2rem 0; border-bottom: 1px solid #ddd;
}
.topic-item:last-child { border-bottom: none; }
.topic-num { font-size: 0.8rem; color: #000; padding-top: 0.1rem; font-weight: 600; }
.topic-content h3 { font-size: 0.92rem; font-weight: 600; margin-bottom: 0.2rem; }
.topic-content p { font-size: 0.85rem; color: #000; line-height: 1.7; }
/* PUBLICATIONS */
#publications { border-top: 1px solid #ddd; }
.pub-list { display: flex; flex-direction: column; }
.pub-item {
padding: 0.9rem 0; border-bottom: 1px solid #ddd;
font-size: 0.85rem; line-height: 1.7; color: #000;
}
.pub-item:last-child { border-bottom: none; }
.pub-item strong { color: #000; font-weight: 500; }
.pub-venue { font-size: 0.72rem; font-weight: 600; color: #b8322e; }
/* RESOURCES */
#resources { border-top: 1px solid #ddd; }
.resources-grid { display: grid; grid-template-columns: 1fr 1fr 1fr; gap: 2rem; }
.resource-item h3 { font-size: 1rem; font-weight: 600; margin-bottom: 0.3rem; }
.resource-item p { font-size: 0.82rem; color: #000; line-height: 1.7; }
.resource-item .badge {
display: inline-block; margin-top: 0.5rem;
font-size: 0.68rem; font-weight: 600; text-transform: uppercase;
letter-spacing: 0.06em; color: #000;
}
/* FOOTER */
footer { border-top: 1px solid #ddd; padding: 2.5rem 2rem; }
footer p { font-size: 0.8rem; color: #000; max-width: 960px; margin: 0 auto; }
footer a { color: #b8322e; text-decoration: none; }
/* MOBILE */
@media (max-width: 768px) {
nav ul { display: none; }
.about-grid, .speakers-grid, .resources-grid { grid-template-columns: 1fr; }
.schedule-item { grid-template-columns: 1fr; gap: 0.5rem; }
.hero-meta { gap: 1.5rem; }
section { padding: 3.5rem 1.2rem; }
}
</style>
</head>
<body>
<!-- NAV -->
<nav>
<div class="inner">
<a href="#" class="logo">UMM <span>Tutorial</span></a>
<ul>
<li><a href="#about">About</a></li>
<li><a href="#schedule">Schedule</a></li>
<li><a href="#speakers">Speakers</a></li>
<li><a href="#topics">Topics</a></li>
<li><a href="#publications">Papers</a></li>
<li><a href="#resources">Resources</a></li>
</ul>
</div>
</nav>
<!-- HERO -->
<section class="hero">
<div class="hero-content">
<div class="hero-badge">CVPR 2026 Tutorial</div>
<h1>The Road to <em>Convergence</em>:<br>Evolution of Unified Multimodal Models</h1>
<p class="subtitle">A comprehensive tutorial on the architecture design, representation learning, training dynamics, and evaluation of unified multimodal models that integrate understanding and generation within a single framework.</p>
<div class="hero-meta">
<div class="item">
<span class="label">Conference</span>
<span class="value highlight">CVPR 2026</span>
</div>
<div class="item">
<span class="label">Duration</span>
<span class="value">Half-Day (~3.5 hrs)</span>
</div>
<div class="item">
<span class="label">Format</span>
<span class="value">Hybrid (In-person + Zoom)</span>
</div>
<div class="item">
<span class="label">Audience</span>
<span class="value">100–300 Attendees</span>
</div>
</div>
</div>
</section>
<!-- ABOUT -->
<section id="about">
<div class="container">
<div class="section-label">Overview</div>
<div class="section-title">Three Central Questions</div>
<div class="section-desc">We structure this tutorial around three fundamental questions that define the design space of unified multimodal models.</div>
<div class="about-grid">
<div class="about-card">
<h3>How to Model?</h3>
<p>A systematic taxonomy of UMM architectures — External Expert Integration, Modular Joint Modeling, and End-to-End Unified Modeling — with trade-off analysis between autoregressive, diffusion, and hybrid approaches.</p>
</div>
<div class="about-card">
<h3>How to Represent?</h3>
<p>The "Unified Tokenizer" debate: continuous representations (e.g., CLIP) vs. discrete tokens (e.g., VQ-VAE), and hybrid encoding strategies balancing semantic understanding with generative fidelity.</p>
</div>
<div class="about-card">
<h3>How to Train?</h3>
<p>The full training lifecycle — from constructing interleaved image-text data to unified pre-training objectives and advanced post-training alignment methods such as DPO and GRPO.</p>
</div>
</div>
</div>
</section>
<!-- SCHEDULE -->
<section id="schedule">
<div class="container">
<div class="section-label">Schedule</div>
<div class="section-title">Tutorial Outline</div>
<div class="section-desc">A structured half-day journey from foundational motivations to advanced architectures and practical training recipes.</div>
<div class="schedule-list">
<div class="schedule-item">
<div class="schedule-time">Session 1<br>30 min</div>
<div class="schedule-body">
<h3>Introduction & Motivation</h3>
<p>Tracing the evolution of multimodal AI from isolated expertise to Unified Multimodal Models. We introduce the core motivations driving unification — particularly the mutual reinforcement between understanding and generation — and provide a rigorous definition of UMMs.</p>
</div>
</div>
<div class="schedule-item">
<div class="schedule-time">Session 2<br>45 min</div>
<div class="schedule-body">
<h3>Modeling Architectures</h3>
<p>A systematic taxonomy including External Expert Integration, Modular Joint Modeling, and End-to-End Unified Modeling. Deep dive into trade-offs between autoregressive, diffusion, and emerging AR-Diffusion hybrid approaches.</p>
</div>
</div>
<div class="schedule-item">
<div class="schedule-time">Session 3<br>45 min</div>
<div class="schedule-body">
<h3>The Unified Tokenizer Challenge</h3>
<p>Comparing continuous representations versus discrete tokenization schemes. Review of encoding/decoding strategies and state-of-the-art hybrid approaches — cascade and dual-branch designs — bridging semantic richness with generative fidelity.</p>
</div>
</div>
<div class="schedule-item">
<div class="schedule-time">Session 4<br>45 min</div>
<div class="schedule-body">
<h3>Training Recipes & Data</h3>
<p>Constructing high-quality modality-interleaved datasets, unified pre-training objectives, and advanced post-training alignment methods including preference-based approaches such as DPO and GRPO.</p>
</div>
</div>
<div class="schedule-item">
<div class="schedule-time">Session 5<br>30 min</div>
<div class="schedule-body">
<h3>Evaluation, Applications & Future Directions</h3>
<p>Reviewing existing benchmarks for standardized evaluation, discussing real-world applications in robotics and autonomous driving, and highlighting open challenges including scalable unified tokenizers and unified world models.</p>
</div>
</div>
<div class="schedule-item">
<div class="schedule-time">Session 6<br>15 min</div>
<div class="schedule-body">
<h3>Unified Codebase & Integration</h3>
<p>A practical walkthrough of our unified multimodal codebase, explaining how core components — tokenizers, multimodal encoders, and generative backbones — are organized and connected in practice.</p>
</div>
</div>
</div>
</div>
</section>
<!-- SPEAKERS -->
<section id="speakers">
<div class="container">
<div class="section-label">Presenters</div>
<div class="section-title">Meet the Team</div>
<div class="section-desc">Our tutorial is led by researchers from academia and industry with extensive experience in multimodal foundation models.</div>
<div class="speakers-grid">
<div class="speaker-card">
<h3>Jindong Wang</h3>
<div class="speaker-role">Assistant Professor, William & Mary</div>
<span class="speaker-tag">Presenter</span>
<div class="speaker-bio">Faculty member of the Future of Life Institute. Former Senior Researcher at Microsoft Research Asia. 60+ papers with 23,000+ citations (h-index 54). World's Top 2% Highly Cited Scientists. Extensive tutorial experience at IJCAI, WSDM, KDD, AAAI, and CVPR.</div>
<a href="https://jd92.wang/" target="_blank" class="speaker-link">jd92.wang →</a>
</div>
<div class="speaker-card">
<h3>Hao Chen</h3>
<div class="speaker-role">Research Scientist, DeepMind</div>
<span class="speaker-tag">Presenter</span>
<div class="speaker-bio">Ph.D. from Carnegie Mellon University (advised by Prof. Bhiksha Raj). Research on data-centric learning for reliable foundation models, including pre-training data imperfections, catastrophic inheritance, and multimodal generalization. Published at NeurIPS, ICML, and ICLR.</div>
</div>
<div class="speaker-card">
<h3>Jiakui Hu</h3>
<div class="speaker-role">Ph.D. Student, Peking University</div>
<span class="speaker-tag">Presenter</span>
<div class="speaker-bio">Research on unified models, computational imaging, and inductive biases in visual foundation models. First-author papers at ICCV, CVPR, ICLR, ICML, and AAAI. Reviewer for major conferences and journals.</div>
</div>
<div class="speaker-card">
<h3>Zhaolong Su</h3>
<div class="speaker-role">Ph.D. Student, William & Mary</div>
<span class="speaker-tag">Presenter</span>
<div class="speaker-bio">Zhaolong Su is a Ph.D. student at William & Mary. He focuses on unified multimodal models, world models, and physical AI. His unified model work includes UniGame (CVPR'26) and FedUMM (WWW'26).</div>
<a href="https://rollingsu.github.io/" target="_blank" class="speaker-link">rollingsu.github.io →</a>
</div>
<div class="speaker-card">
<h3>Sharon Li</h3>
<div class="speaker-role">Associate Professor, UW–Madison</div>
<span class="speaker-tag">Advisor</span>
<div class="speaker-bio">Research on reliable and safe AI systems. Alfred P. Sloan Fellowship and MIT Technology Review Innovators Under 35 recipient. Ph.D. from Cornell University, postdoc at Stanford University.</div>
</div>
</div>
</div>
</section>
<!-- TOPICS -->
<section id="topics">
<div class="container">
<div class="section-label">Coverage</div>
<div class="section-title">Key Topics</div>
<div class="section-desc">From architectural paradigms to real-world deployment, the tutorial covers the full spectrum of unified multimodal model research.</div>
<div class="topics-list">
<div class="topic-item">
<div class="topic-num">01</div>
<div class="topic-content">
<h3>Evolution of Multimodal Models</h3>
<p>From isolated multimodal understanding or generation systems to unified multimodal foundation models capable of handling both tasks simultaneously.</p>
</div>
</div>
<div class="topic-item">
<div class="topic-num">02</div>
<div class="topic-content">
<h3>Modeling Paradigms for UMMs</h3>
<p>A taxonomy of architectures including External Expert Integration, Modular Joint Modeling, and End-to-End Unified Modeling, with comparisons between autoregressive, diffusion, and hybrid approaches.</p>
</div>
</div>
<div class="topic-item">
<div class="topic-num">03</div>
<div class="topic-content">
<h3>Unified Tokenizer & Representation Design</h3>
<p>Continuous versus discrete representations, their advantages and limitations, and emerging hybrid encoding strategies that balance semantic understanding and generative fidelity.</p>
</div>
</div>
<div class="topic-item">
<div class="topic-num">04</div>
<div class="topic-content">
<h3>Training Lifecycle & Alignment</h3>
<p>Construction of modality-interleaved datasets, unified pre-training objectives, and post-training alignment methods such as DPO and GRPO.</p>
</div>
</div>
<div class="topic-item">
<div class="topic-num">05</div>
<div class="topic-content">
<h3>Benchmarks, Applications & Open Challenges</h3>
<p>Evaluation protocols, real-world applications in robotics and autonomous driving, and future directions such as scalable unified tokenizers and unified world models.</p>
</div>
</div>
</div>
</div>
</section>
<!-- PUBLICATIONS -->
<section id="publications">
<div class="container">
<div class="section-label">Related Work</div>
<div class="section-title">Selected Publications</div>
<div class="section-desc">Representative publications by the organizers and foundational research in unified multimodal models.</div>
<div class="pub-list">
<div class="pub-item">
<strong>Jiakui Hu, et al.</strong> Unified Multimodal Understanding and Generation Models: Advances, Challenges, and Opportunities. <span class="pub-venue">Survey</span>
</div>
<div class="pub-item">
<strong>Jindong Wang, Hao Chen, et al.</strong> On Fairness of Unified Multimodal Large Language Models for Image Generation. <span class="pub-venue">NeurIPS 2025</span>
</div>
<div class="pub-item">
<strong>Jindong Wang, Hao Chen, et al.</strong> Is Your (Reasoning) Multimodal Language Model Vulnerable toward Distractions? <span class="pub-venue">AAAI 2026</span>
</div>
<div class="pub-item">
<strong>Hao Chen, et al.</strong> ImageFolder: Autoregressive Image Generation with Folded Tokens. <span class="pub-venue">ICLR 2025</span>
</div>
<div class="pub-item">
<strong>Hao Chen, et al.</strong> Masked Autoencoders Are Effective Tokenizers for Diffusion Models. <span class="pub-venue">ICML 2025</span>
</div>
<div class="pub-item">
<strong>Sharon Li, et al.</strong> Understanding Multimodal LLMs Under Distribution Shifts: An Information-Theoretic Approach. <span class="pub-venue">ICML 2025 Oral</span>
</div>
<div class="pub-item">
<strong>Jindong Wang, et al.</strong> Open-Vocabulary Calibration for Vision–Language Models. <span class="pub-venue">ICML 2024</span>
</div>
<div class="pub-item">
<strong>Zhaolong Su, Hao Chen, Jindong Wang, et al.</strong> UniGame: Turning a Unified Multimodal Model Into Its Own Adversary. <span class="pub-venue">CVPR 2026</span>
</div>
</div>
</div>
</section>
<!-- RESOURCES -->
<section id="resources">
<div class="container">
<div class="section-label">Open Science</div>
<div class="section-title">Materials & Resources</div>
<div class="section-desc">We are committed to open science and ensuring reproducibility. All materials will be publicly available.</div>
<div class="resources-grid">
<div class="resource-item">
<h3>Slides</h3>
<p>All presentation slides will be made publicly available on this website following the event.</p>
<span class="badge">Coming Soon</span>
</div>
<div class="resource-item">
<h3>Bibliography</h3>
<p>An annotated compilation of all references discussed in the tutorial as a comprehensive reading list.</p>
<span class="badge">Coming Soon</span>
</div>
<div class="resource-item">
<h3>Codebase</h3>
<p>Open-source unified multimodal codebase with annotated pointers to models (e.g., Emu, Janus) and datasets.</p>
<span class="badge">Coming Soon</span>
</div>
</div>
</div>
</section>
<!-- FOOTER -->
<footer>
<p>CVPR 2026 Tutorial — The Road to Convergence · <a href="https://umm-tutorial.github.io" target="_blank">umm-tutorial.github.io</a></p>
</footer>
</body>
</html>