-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
592 lines (544 loc) · 30.1 KB
/
index.html
File metadata and controls
592 lines (544 loc) · 30.1 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
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>MIND: Benchmarking Memory Consistency and Action Control in World Models</title>
<link rel="shortcut icon" href="resources/icon/logo_square.png">
<!-- Fonts and Icons
<link rel="preconnect" href="https://fonts.googleapis.com" /> -->
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin />
<link href='https://fonts.googleapis.com/css?family=Titillium+Web:400,600,400italic,600italic,300,300italic' rel='stylesheet' type='text/css'>
<link href="https://fonts.googleapis.com/icon?family=Material+Icons" rel="stylesheet" />
<!-- Main stylesheet reused from the reference project -->
<link href="resources/style.css" rel="stylesheet" type="text/css" />
</head>
<body>
<br /><br />
<center>
<span class="header-text" style="display: inline-block; margin-bottom: 5px;"> <!-- 关键：减少底部间距 -->
<img src="resources/icon/logo_transparent.png" alt="logo" style="height: 1em; vertical-align: top;"/> <span style="color:#1F82C0">M</span><span style="color:#1CBF91">I</span><span style="color:#39C46E">N</span><span style="color:#149C7E">D</span>: Benchmarking Memory Consistency and Action Control <br> in World Models
</span>
</center>
<center>
<h6 class="tldr-section" style="color: red; margin-top: 0; margin-bottom: 0;"> <!-- 关键：移除顶部默认间距 -->
TL;DR: The first open-domain closed-loop revisited benchmark for evaluating memory consistency and action control in world models
</h6>
</center>
<center>
<div class="author-block" style="font-size: 20px; line-height: 1.6; margin: 10px 0;">
<!-- 作者名称及上标标注 -->
<div style="margin-bottom: 15px;">
<b>
<a href="https://easontut.github.io/" target="_blank" style="text-decoration: none; color: inherit;">Yixuan Ye</a><sup>1*</sup>,
<a href="" target="_blank" style="text-decoration: none; color: inherit;">Xuanyu Lu</a><sup>1*</sup>,
<a href="https://yuxinn-j.github.io/" target="_blank" style="text-decoration: none; color: inherit;">Yuxin Jiang</a><sup>2*</sup>,
<a href="https://ycgu.site/" target="_blank" style="text-decoration: none; color: inherit;">Yuchao Gu</a><sup>2</sup>,
<a href="https://ruizhaocv.github.io/" target="_blank" style="text-decoration: none; color: inherit;">Rui Zhao</a><sup>2</sup>,
<a href="https://kolakivy.github.io/" target="_blank" style="text-decoration: none; color: inherit;">Qiwei Liang</a><sup>3</sup>,
<a href="" target="_blank" style="text-decoration: none; color: inherit;">Jiachun Pan</a><sup>2</sup>,
<br style="line-height: inherit;"/>
<a href="" target="_blank" style="text-decoration: none; color: inherit;">Fengda Zhang</a><sup>4</sup>,
<a href="https://weijiawu.github.io/" target="_blank" style="text-decoration: none; color: inherit;">Weijia Wu</a><sup>2†</sup>,
<a href="https://fingerrec.github.io/" target="_blank" style="text-decoration: none; color: inherit;">Alex Jinpeng Wang</a><sup>1†</sup>
</b>
</div>
<!-- 机构标注行1 -->
<div style="font-size: 20px; margin-bottom: 8px;">
<sup>1</sup> CSU-JPG, Central South University
<sup>2</sup> National University of Singapore
</div>
<!-- 机构标注行2 -->
<div style="font-size: 20px;">
<sup>3</sup> Hong Kong University of Science and Technology (Guangzhou)
<sup>4</sup> Nanyang Technological University
</div>
</div>
</center>
<center>
<!-- 优化后的导航链接区域（带悬浮特效） -->
<div style="margin: 20px 0; font-size: 20px; font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif; text-align: center;">
<a href="https://csu-jpg.github.io/MIND.github.io/" style="margin: 0 12px; text-decoration: none; color: #333; transition: all 0.3s ease; display: inline-block;">
🌐 Homepage
</a> |
<a href="https://huggingface.co/datasets/CSU-JPG/MIND" style="margin: 0 12px; text-decoration: none; color: #333; transition: all 0.3s ease; display: inline-block;">
👉 Dataset
</a> |
<a href="https://arxiv.org/abs/2602.08025" style="margin: 0 12px; text-decoration: none; color: #333; transition: all 0.3s ease; display: inline-block;">
📄 Paper
</a> |
<a href="https://github.com/CSU-JPG/MIND" style="margin: 0 12px; text-decoration: none; color: #333; transition: all 0.3s ease; display: inline-block;">
💻 Code
</a> |
<a href="https://csu-jpg.github.io/MIND.github.io/" style="margin: 0 12px; text-decoration: none; color: #333; transition: all 0.3s ease; display: inline-block;">
🏆 Leaderboard (coming soon)
</a>
<style>
/* 悬浮特效样式 */
div a:hover {
color: #1F82C0; /* 匹配MIND中M的蓝色，视觉更统一 */
font-weight: 600; /* 文字加粗 */
transform: scale(1.05); /* 轻微放大 */
text-decoration: underline; /* 下划线（可选） */
}
</style>
</div>
<div class="image-container">
<img id="teaser-image" src="resources/intro/first_figure.png" alt="Project Teaser" />
</div>
</center>
<!-- Abstract -->
<div class="docs-section" id="abstract">
<center>
<h1 class="navbar-heading"><b>Abstract</b></h1>
</center>
<div class="text-container">
<p>
World models aim to understand, remember, and predict dynamic visual environments, yet a unified benchmark for evaluating their fundamental abilities remains lacking. To address this gap, we introduce
<span style="color:#1F82C0">M</span><span style="color:#1CBF91">I</span><span style="color:#39C46E">N</span><span style="color:#149C7E">D</span>, the first open-domain benchmark for evaluating <span style="color:#1F82C0">M</span>emory cons<span style="color:#1CBF91">I</span>stency and action co<span style="color:#39C46E">N</span>trol in worl<span style="color:#149C7E">D</span> models. <span style="color:#1F82C0">M</span><span style="color:#1CBF91">I</span><span style="color:#39C46E">N</span><span style="color:#149C7E">D</span> contains 250 high-quality videos at 1080p and 24 FPS, including 100 (first-person) + 100 (third-person) video clips under a shared action space and 25 + 25 clips across varied action spaces covering eight diverse scenes. We design an efficient evaluation framework to measure two core abilities: memory consistency and action control, capturing temporal stability and contextual coherence across viewpoints. Furthermore, we design various action spaces, including different character movement speeds and camera rotation angles, to evaluate the action generalization capability across different action spaces under shared scenes. To facilitate future performance benchmarking on <span style="color:#1F82C0">M</span><span style="color:#1CBF91">I</span><span style="color:#39C46E">N</span><span style="color:#149C7E">D</span>, we introduce MIND-World, a novel interactive Video-to-World baseline. Extensive experiments demonstrate the completeness of <span style="color:#1F82C0">M</span><span style="color:#1CBF91">I</span><span style="color:#39C46E">N</span><span style="color:#149C7E">D</span> and reveal key challenges in current world models, including the difficulty of maintaining long-term memory consistency and generalizing across action spaces.
</p>
</div>
</div>
<!-- Method / Overview -->
<div class="docs-section" id="method">
<center>
<h1 class="navbar-heading"><b>Overview</b></h1>
</center>
<div class="image-container">
<img id="overview-image" src="resources/intro/overview.png" alt="HAMLET Overview Diagram" />
</div>
<div class="text-container">
<p>
Overview of the <span style="color:#1F82C0">M</span><span style="color:#1CBF91">I</span><span style="color:#39C46E">N</span><span style="color:#149C7E">D</span>. We build and collect the first open-domain benchmark using Unreal Engine 5, supporting both first-person and third-person perspectives with 1080p resolution at 24 FPS.
</p>
</div>
</div>
<!-- Method / Overview -->
<div class="docs-section" id="method">
<center>
<h1 class="navbar-heading"><b><span style="color:#1F82C0">M</span><span style="color:#1CBF91">I</span><span style="color:#39C46E">N</span><span style="color:#149C7E">D</span> Dataset</b>
<button class="play-all-btn" onclick="toggleVideoGroup(this)">
<img class="play-all-icon" src="resources/icon/play.svg" alt="Play All" />
</button>
</h1>
</center>
<div class="video-item">
<video src="resources/intro/scene_video.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="text-container">
<p>
Distribution for Scene Categories and Action Space in <span style="color:#1F82C0">M</span><span style="color:#1CBF91">I</span><span style="color:#39C46E">N</span><span style="color:#149C7E">D</span> Dataset. <span style="color:#1F82C0">M</span><span style="color:#1CBF91">I</span><span style="color:#39C46E">N</span><span style="color:#149C7E">D</span> supports open-domain scenarios with diverse and well-balanced action spaces.
</p>
</div>
</div>
<!--
<div class="docs-section" id="results">
<center>
<h1 class="navbar-heading">
<b>Scene</b>
</h1>
</center>
<div class="image-item">
<img src="resources/intro/scene.png" alt="Challenge 2 Diagram">
<div class="text-container">
<p>hello world</p>
</div>
</div>
</div> -->
<!-- Results -->
<div class="docs-section" id="results">
<center>
<div class="navbar-heading-demo">
<div class="title-row">
<b style="font-size: 2rem;">Demo</b>
<button class="play-all-btn" onclick="toggleVideoGroup(this)">
<img class="play-all-icon" src="resources/icon/play.svg" alt="Play All" />
</button>
</div>
<p class="subtitle-hint">
(The closer the inferred video is to the ground truth, the better.)
</p>
</div>
</center>
<div class="task-block" id="task-1">
<center><h3 style="font-size: 1.35em; margin: 0.8em 0; font-weight: 700;">Ground Truth</h3></center>
<div class="video-grid four-col-grid" aria-label="Task 1-a">
<div class="video-item">
<video src="resources/demo1/2025-10-30-17-58-LandscapePackOne_SnowyMountain01-3m21s-gt.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo2/2025-11-5-21-24-Paragon_Sample-0m33s-gt.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo3/2025-10-18-20-24-HandPaintedEnvironment_Winter_Scene-2m06s-gt.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo4/2025-10-20-20-3-Industrial_Warehouse-1m31s-gt.mp4" playsinline muted loop preload="auto"></video>
</div>
</div>
<center><h3 style="font-size: 1.35em; margin: 0.8em 0; font-weight: 700;">MIND-World (Ours) <i style="color:#00D084;">w/ Memory</i></h3></center>
<div class="video-grid four-col-grid" aria-label="Task 1-a">
<div class="video-item">
<video src="resources/demo1/2025-10-30-17-58-LandscapePackOne_SnowyMountain01-3m21s-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo2/2025-11-5-21-24-Paragon_Sample-0m33s-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo3/2025-10-18-20-24-HandPaintedEnvironment_Winter_Scene-2m06s-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo4/2025-10-20-20-3-Industrial_Warehouse-1m31s-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
</div>
<center><h3 style="font-size: 1.35em; margin: 0.8em 0; font-weight: 700;">MIND-World (Ours) <i style="color:#c90f0f;">w/o Memory</i></h3></center>
<div class="video-grid four-col-grid" aria-label="Task 1-a">
<div class="video-item">
<video src="resources/demo1/2025-10-30-17-58-LandscapePackOne_SnowyMountain01-3m21s-i2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo2/2025-11-5-21-24-Paragon_Sample-0m33s-i2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo3/2025-10-18-20-24-HandPaintedEnvironment_Winter_Scene-2m06s-i2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo4/2025-10-20-20-3-Industrial_Warehouse-1m31s-i2v.mp4" playsinline muted loop preload="auto"></video>
</div>
</div>
<center><h3 style="font-size: 1.35em; margin: 0.8em 0; font-weight: 700;">Matrix-game-2.0</h3></center>
<div class="video-grid four-col-grid" aria-label="Task 1-a">
<div class="video-item">
<video src="resources/demo1/2025-10-30-17-58-LandscapePackOne_SnowyMountain01-3m21s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo2/2025-11-5-21-24-Paragon_Sample-0m33s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo3/2025-10-18-20-24-HandPaintedEnvironment_Winter_Scene-2m06s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/demo4/2025-10-20-20-3-Industrial_Warehouse-1m31s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
</div>
</div>
</div>
<div class="docs-section" id="results">
<center>
<h1 class="navbar-heading">
<b>Challenge 1: Open Domain</b>
<button class="play-all-btn" onclick="toggleVideoGroup(this)">
<img class="play-all-icon" src="resources/icon/play.svg" alt="Play All" />
</button>
</h1>
</center>
<div class="task-block" id="task-1">
<div class="video-grid four-col-grid" aria-label="Task 1-a">
<div class="video-item">
<div class="video-caption">Origin</div>
<img src="resources/challenge1/out_gt.png" alt="Ground Truth First Frame"/>
</div>
<div class="video-item">
<div class="video-caption">Trained on <span style="color:#1F82C0">M</span><span style="color:#1CBF91">I</span><span style="color:#39C46E">N</span><span style="color:#149C7E">D</span> Dataset</Data></div>
<video src="resources/challenge1/out-mind.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<div class="video-caption">Trained on Minecraft Dataset</div>
<video src="resources/challenge1/out-minecraft.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="text-container" style="display: flex; align-items: center; justify-content: center; height: 100%;">
<p>How to achieve open-domain generalization with readily available large-scale data (e.g., Minecraft Datasets)?</p>
</div>
</div>
</div>
</div>
<div class="docs-section" id="results">
<!-- <center>
<h1 class="navbar-heading">
<b>Challenge 2: Action Space Generalization</b>
<button class="play-all-btn" onclick="toggleVideoGroup(this)">
<img class="play-all-icon" src="resources/icon/play.svg" alt="Play All" />
</button>
</h1>
</center> -->
<div class="navbar-heading-demo">
<div class="title-row">
<b style="font-size: 2rem;">Challenge 2: Action Space Generalization</b>
<button class="play-all-btn" onclick="toggleVideoGroup(this)">
<img class="play-all-icon" src="resources/icon/play.svg" alt="Play All" />
</button>
</div>
<p class="subtitle-hint">
(The closer the inferred video is to the ground truth, the better.)
</p>
</div>
<div class="video-grid two-col-grid" aria-label="Task 1-a">
<div class="video-grid one-three-col-grid" aria-label="Task 1-a">
<div class="text-container" style="display: flex; align-items: center; justify-content: center; height: 100%;">
<p style="font-size: 24px; font-weight: bold; margin: 0;">0.8 X</p>
</div>
<div class="video-item">
<div class="video-caption"><i style="color:#c90f0f;">w/o Memory</i></div>
<video src="resources/challenge2/2025-11-8-18-37-Downtown_West-0.8x-175-4m35s-i2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<div class="video-caption">Ground Truth</div>
<video src="resources/challenge2/2025-11-8-18-37-Downtown_West-0.8x-175-4m35s-gt.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<div class="video-caption"><i style="color:#00D084;">w/ Memory</i></div>
<video src="resources/challenge2/2025-11-8-18-37-Downtown_West-0.8x-175-4m35s-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="text-container" style="display: flex; align-items: center; justify-content: center; height: 100%;">
<p style="font-size: 24px; font-weight: bold; margin: 0;">1.0 X</p>
</div>
<div class="video-item">
<video src="resources/challenge2/2025-10-30-14-3-Cyberpunk_Industries-0m23s-i2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/challenge2/2025-10-30-14-3-Cyberpunk_Industries-0m23s-gt.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/challenge2/2025-10-30-14-3-Cyberpunk_Industries-0m23s-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
</div>
<div class="image-item">
<img src="resources/challenge2/challenge_2.png" alt="Challenge 2 Diagram">
</div>
</div>
<div class="text-container">
<p>
How to overcome the interference caused by the mismatch between the implicit action space in context memory and the training distribution, thereby preventing catastrophic performance degradation of memory-augmented models during action space generalization?
</p>
</div>
</div>
<div class="docs-section" id="results">
<center>
<h1 class="navbar-heading">
<b>Challenge 3: Precise Action Control</b>
<button class="play-all-btn" onclick="toggleVideoGroup(this)">
<img class="play-all-icon" src="resources/icon/play.svg" alt="Play All" />
</button>
</h1>
</center>
<center><h3 style="font-size: 1.35em; margin: 0.8em 0; font-weight: 700;">MIND-World (Ours)</h3></center>
<div class="video-grid four-col-grid" aria-label="Task 1-a">
<div class="video-item">
<video src="resources/challenge3/2025-10-18-16-31-BoardcastStudio_A-1m01s_-a_idx428_s0-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/challenge3/2025-10-18-19-59-HandPaintedEnvironment_Winter_Scene-1m31s_-wr_idx227_s0-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/challenge3/2025-10-19-0-33-Stylized_Egypt-4m31s_-wl_idx41_s0-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/challenge3/2025-10-19-15-44-Starter_Bundle-4m43s_-a_idx353_s0-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
</div>
<center><h3 style="font-size: 1.35em; margin: 0.8em 0; font-weight: 700;">Matrix-game-2.0</h3></center>
<div class="video-grid four-col-grid" aria-label="Task 1-a">
<div class="video-item">
<video src="resources/challenge3/2025-10-22-23-40-DowntownWest-4m58s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/challenge3/2025-10-20-19-43-Industrial_Warehouse-2m08s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/challenge3/2025-10-30-22-35-Stunning_Level-3m38s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/challenge3/2025-11-5-17-4-Factory_Environment-4m12s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
</div>
<div class="text-container">
<p>
How to effectively decouple visual prompts from action dynamics, thereby preventing world models from being interfered by initial images/videos when executing precise action sequences, and achieving true precise action control?
</p>
</div>
</div>
<div class="docs-section" id="results">
<!-- <center>
<h1 class="navbar-heading">
<b>Challenge 4: Long Context Memory</b>
<button class="play-all-btn" onclick="toggleVideoGroup(this)">
<img class="play-all-icon" src="resources/icon/play.svg" alt="Play All" />
</button>
</h1>
</center> -->
<div class="navbar-heading-demo">
<div class="title-row">
<b style="font-size: 2rem;">Challenge 4: Long Context Memory</b>
<button class="play-all-btn" onclick="toggleVideoGroup(this)">
<img class="play-all-icon" src="resources/icon/play.svg" alt="Play All" />
</button>
</div>
<p class="subtitle-hint">
(The closer the inferred video is to the ground truth, the better.)
</p>
</div>
<div class="video-grid two-col-grid" aria-label="Task 1-a">
<div class="video-grid three-col-grid" aria-label="Task 1-a">
<div class="video-item">
<div class="video-caption"><i style="color:#c90f0f;">w/o Memory</i></div>
<video src="resources/challenge4/2025-10-20-18-16-Stylized_Kitchen-0m54s-i2v.mp4" playsinline muted preload="auto"></video>
</div>
<div class="video-item">
<div class="video-caption">Ground Truth</div>
<video src="resources/challenge4/2025-10-20-18-16-Stylized_Kitchen-0m54s-gt.mp4" playsinline muted preload="auto"></video>
</div>
<div class="video-item">
<div class="video-caption"><i style="color:#00D084;">w/ Memory</i></div>
<video src="resources/challenge4/2025-10-20-18-16-Stylized_Kitchen-0m54s-v2v.mp4" playsinline muted preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/challenge4/2025-10-18-18-39-HandPaintedEnvironment_Summer_Scene-3m31s-i2v.mp4" playsinline muted preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/challenge4/2025-10-18-18-39-HandPaintedEnvironment_Summer_Scene-3m31s-gt.mp4" playsinline muted preload="auto"></video>
</div>
<div class="video-item">
<video src="resources/challenge4/2025-10-18-18-39-HandPaintedEnvironment_Summer_Scene-3m31s-v2v.mp4" playsinline muted preload="auto"></video>
</div>
</div>
<div class="image-item">
<img src="resources/challenge4/challenge4.png" />
</div>
</div>
<div class="text-container">
<p>
How to effectively maintain and leverage long-context memory, thereby enabling world models to achieve and sustain high consistency with ground-truth trajectories in ultra-long-horizon rollouts?
</p>
</div>
</div>
<div class="docs-section" id="results">
<center>
<h1 class="navbar-heading">
<b>Challenge 5: Generated Scene Consistency</b>
<button class="play-all-btn" onclick="toggleVideoGroup(this)">
<img class="play-all-icon" src="resources/icon/play.svg" alt="Play All" />
</button>
</h1>
</center>
<div class="video-grid four-col-grid" aria-label="Task 1-a">
<div class="video-item">
<div class="video-caption">Matrix-game-2.0</div>
<video src="resources/challenge5/2025-10-18-18-39-HandPaintedEnvironment_Summer_Scene-3m31s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<div class="video-caption">Matrix-game-2.0</div>
<video src="resources/challenge5/2025-10-19-0-33-Stylized_Egypt-4m31s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<div class="video-caption">MIND-World (Ours)</div>
<video src="resources/challenge5/2025-10-19-12-49-ObsidianGoldOrnates_YV-1m46s_-sl_idx54_s0-i2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<div class="video-caption">MIND-World (Ours)</div>
<video src="resources/challenge5/2025-10-18-21-12-Namaqualand-4m05s_-sl_idx327_s0-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
</div>
<div class="text-container">
<p>
How to ensure high consistency in world models when repeatedly generating the same region?
</p>
</div>
</div>
<div class="docs-section" id="results">
<center>
<h1 class="navbar-heading">
<b>Challenge 6: Third-person Perspective</b>
<button class="play-all-btn" onclick="toggleVideoGroup(this)">
<img class="play-all-icon" src="resources/icon/play.svg" alt="Play All" />
</button>
</h1>
</center>
<div class="video-grid four-col-grid" aria-label="Task 1-a">
<div class="video-item">
<div class="video-caption">Matrix-game-2.0</div>
<video src="resources/challenge6/2025-10-30-15-54-Downtown_West-4m15s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<div class="video-caption">Matrix-game-2.0</div>
<video src="resources/challenge6/2025-11-2-21-34-Soul_City-0m57s-matrix_game.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<div class="video-caption">MIND-World (Ours)</div>
<video src="resources/challenge6/2025-10-30-13-1-Fantastic_Village-5m06s-i2v.mp4" playsinline muted loop preload="auto"></video>
</div>
<div class="video-item">
<div class="video-caption">MIND-World (Ours)</div>
<video src="resources/challenge6/2025-11-3-16-46-CityKit_BR-3m09s-v2v.mp4" playsinline muted loop preload="auto"></video>
</div>
</div>
<div class="text-container">
<p>
How to endow world models with the ability to correctly model physical interactions and spatial relationships between agents and the environment?
</p>
</div>
</div>
<!-- BibTeX 引用区块（仅修改内容，样式/功能不变） -->
<div class="docs-section" id="citation" style="background-color: #f8f9fa; border-radius: 8px; padding: 25px; margin: 40px auto; max-width: 1200px; border: 1px solid #e9ecef;">
<center>
<h1 class="navbar-heading"><b>Citation</b></h1>
</center>
<div style="position: relative; margin-top: 20px;">
<!-- 复制按钮（样式/功能不变） -->
<button id="copyBibtexBtn" style="position: absolute; top: 10px; right: 10px; background-color: #1F82C0; color: white; border: none; border-radius: 4px; padding: 6px 12px; cursor: pointer; font-size: 14px; display: flex; align-items: center; gap: 6px; transition: background-color 0.2s;">
<span class="material-icons" style="font-size: 18px;">content_copy</span>
Copy
</button>
<!-- 仅修改这里的BibTeX内容即可 -->
<pre id="bibtexCode" style="background-color: #f1f3f5; border-radius: 6px; padding: 20px; font-family: 'Consolas', 'Monaco', monospace; font-size: 14px; line-height: 1.6; overflow-x: auto; margin: 0;">
@misc{ye2026mind,
title={MIND: Benchmarking Memory Consistency and Action Control in World Models},
author={Yixuan Ye and Xuanyu Lu and Yuxin Jiang and Yuchao Gu and Rui Zhao and Qiwei Liang and Jiachun Pan and Fengda Zhang and Weijia Wu and Alex Jinpeng Wang},
year={2026},
eprint={2602.08025},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2602.08025},
}
</pre>
</div>
<p style="margin-top: 15px; font-size: 16px; color: #495057; text-align: center;">
If you find our work useful, please consider citing our paper.
</p>
</div>
<!-- 原有脚本无需修改（复制功能保持不变） -->
<script>
// 复制BibTeX功能
document.getElementById('copyBibtexBtn').addEventListener('click', function() {
const bibtexText = document.getElementById('bibtexCode').textContent.trim();
navigator.clipboard.writeText(bibtexText).then(() => {
const originalText = this.innerHTML;
this.innerHTML = '<span class="material-icons" style="font-size: 18px;">check</span> Copied!';
this.style.backgroundColor = '#1CBF91';
setTimeout(() => {
this.innerHTML = originalText;
this.style.backgroundColor = '#1F82C0';
}, 2000);
}).catch(err => {
alert('Failed to copy: ' + err);
});
});
// 原有视频播放控制函数
function toggleVideoGroup(button) {
const icon = button.querySelector('.play-all-icon');
const videoContainer = button.closest('.docs-section');
if (!videoContainer || !icon) return;
const videos = videoContainer.querySelectorAll('video');
const isPlaying = icon.src.includes('pause.svg');
if (isPlaying) {
videos.forEach(video => video.pause());
icon.src = 'resources/icon/play.svg';
icon.alt = 'Play All';
} else {
videos.forEach(video => video.play());
icon.src = 'resources/icon/pause.svg';
icon.alt = 'Pause All';
}
}
</script>
</body>
</html>