-
Notifications
You must be signed in to change notification settings - Fork 3
Expand file tree
/
Copy pathindex.html
More file actions
632 lines (588 loc) · 34.5 KB
/
index.html
File metadata and controls
632 lines (588 loc) · 34.5 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
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
<!DOCTYPE html>
<html>
<head>
<meta charset="UTF-8">
<title>EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents</title>
<!-- Favicon -->
<link rel="icon" href="website/img/embodied-logo.png" type="image/png">
<meta name="viewport" content="width=device-width, initial-scale=1">
<script src="https://cdn.jsdelivr.net/npm/chart.js@4.4.0/dist/chart.umd.min.js"></script>
<script src="https://cdn.jsdelivr.net/npm/chartjs-plugin-datalabels@2.0.0"></script>
<script
src="https://cdn.jsdelivr.net/npm/chartjs-plugin-annotation@3.0.1/dist/chartjs-plugin-annotation.min.js"></script>
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
<link rel="stylesheet" href="website/css/bulma.min.css">
<link rel="stylesheet" href="website/css/bulma-carousel.min.css">
<link rel="stylesheet" href="website/css/bulma-slider.min.css">
<link rel="stylesheet" href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="website/css/fontawesome.all.min.css">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script src="./website/javascript/bulma-carousel.min.js"></script>
<script src="./website/javascript/bulma-slider.min.js"></script>
<script src="./website/javascript/explorer-index.js"></script>
<link href="https://cdn.jsdelivr.net/npm/bootstrap@5.1.3/dist/css/bootstrap.min.css" rel="stylesheet"
integrity="sha384-1BmE4kWBq78iYhFldvKuhfTAU6auU8tT94WrHftjDbrCEXSU1oBoqyl2QvZ6jIW3" crossorigin="anonymous">
<script src="https://cdn.jsdelivr.net/npm/bootstrap@5.1.3/dist/js/bootstrap.bundle.min.js"
integrity="sha384-ka7Sk0Gln4gmtz2MlQnikT1wXgYsOg+OMhuP+IlRH9sENBO0LRn5q+8nbTov4+1p"
crossorigin="anonymous"></script>
<link href="https://unpkg.com/tabulator-tables@5.5.2/dist/css/tabulator_bootstrap4.min.css" rel="stylesheet">
<script type="text/javascript" src="https://unpkg.com/tabulator-tables@5.5.2/dist/js/tabulator.min.js"></script>
<script defer src="website/javascript/fontawesome.all.min.js"></script>
<!-- <script src="website/javascript/peity-vanilla.js"></script> -->
<!-- below we load some js scripts -->
<script src="website/javascript/benchmark_table.js" type="module"></script>
<script src="website/javascript/benchmark_table_low_level.js" type="module"></script>
<link rel="stylesheet" href="website/css/index.css">
<!-- Google tag (gtag.js) -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-C7GJ4FYMY9"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag() { dataLayer.push(arguments); }
gtag('js', new Date());
gtag('config', 'G-C7GJ4FYMY9');
</script>
<!-- MathJax script -->
<script type="text/javascript" async
src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.5/MathJax.js?config=TeX-MML-AM_CHTML">
</script>
<script type="text/javascript">
MathJax.Hub.Config({
tex2jax: {inlineMath: [['$','$'], ['\\(','\\)']]}
});
</script>
<noscript>
<p><img alt="Clicky" width="1" height="1" src="//in.getclicky.com/101339888ns.gif" /></p>
</noscript>
<script>
document.addEventListener('DOMContentLoaded', function () {
var toggles = document.querySelectorAll('.toggle-section');
toggles.forEach(function(toggle) {
toggle.addEventListener('click', function() {
var content = document.getElementById(toggle.getAttribute('aria-controls'));
var toggleIcon = toggle.children[1].children[0];
content.classList.toggle('is-active');
if (content.classList.contains('is-active')) {
toggleIcon.style.transition = 'transform 0.3s ease';
toggleIcon.style.transform = 'rotate(180deg)';
} else {
toggleIcon.style.transition = 'transform 0.3s ease';
toggleIcon.style.transform = 'rotate(0deg)';
}
});
});
});
</script>
<style>
.collapse-content {
display: none;
margin-top: 10px;
}
.collapse-content.is-active {
display: block;
}
/* .toggle-section .icon.is-small {
transition: transform 0.3s ease;
} */
/* .toggle-section .fa-angle-up {
transform: rotate(180deg);
} */
</style>
</head>
<body>
<!-- Navigation Bar -->
<nav class="navbar" role="navigation" aria-label="main navigation" style="background: #fff; box-shadow: 0 1px 3px rgba(0,0,0,0.1);">
<div class="container">
<div class="navbar-brand">
<a class="navbar-item" href="index.html">
<img src="website/img/embodied-logo.png" alt="EmbodiedBench" width="40" height="40">
<span class="embodiedbench" style="margin-left: 0.5rem; font-size: 1.25rem;">EmbodiedBench</span>
</a>
</div>
<div class="navbar-end">
<a href="index.html" class="navbar-item has-text-weight-bold">Home</a>
<a href="challenge.html" class="navbar-item" style="color: #5e7154;">Challenge</a>
</div>
</div>
</nav>
<!-- Single-line Title with Bulma's .title.is-1 is-bold (centered heading) -->
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<!-- MAIN TITLE (logo + "EmbodiedBench:") -->
<h1 class="title is-1 publication-title is-bold">
<img src="website/img/embodied-logo.png"
alt="logo"
style="vertical-align: middle; margin-right: 20px; position: relative; top: -15px; left: 30px"
width="80"
height="80" />
<span class="embodiedbench" style="vertical-align: middle">EmbodiedBench</span>
</h1>
<!-- SUBTITLE -->
<h2 class="title is-1 publication-title is-bold" style="margin-top:-0.7em; margin-right: 30px; margin-left: 30px;">
Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents
</h2>
<!-- Author list -->
<div class="is-size-5 publication-authors">
<span class="author-block">
<a href="https://yangrui2015.github.io/" target="_blank">Rui Yang</a><sup>1*</sup>,
</span>
<span class="author-block">
<a href="https://openreview.net/profile?id=~Hanyang_Chen3" target="_blank">Hanyang Chen</a><sup>1*</sup>,
</span>
<span class="author-block">
<a href="https://jyzhang1208.github.io/" target="_blank">Junyu Zhang</a><sup>1*</sup>,
</span>
<span class="author-block">
<a href="https://www.linkedin.com/in/markzihaozhao/" target="_blank">Mark Zhao</a><sup>3*</sup>,
</span><br>
<span class="author-block">
<a href="https://qiancheng0.github.io/">Cheng Qian</a><sup>1</sup>,
</span>
<span class="author-block">
<a href="https://jameskrw.github.io/">Kangrui Wang</a><sup>2</sup>,
</span>
<span class="author-block">
<a href="https://qinengwang-aiden.github.io/">Qineng Wang</a><sup>2</sup>,
</span>
<span class="author-block">
<a href="https://www.linkedin.com/in/teja-koripella/">Teja Venkat Koripella</a><sup>1</sup>,
</span>
<span class="author-block">
<a href="https://www.linkedin.com/in/marziyeh-movahedi-439b7284/" target="_blank">Marziyeh Movahedi</a><sup>4</sup>,
</span><br>
<span class="author-block">
<a href="https://limanling.github.io/">Manling Li</a><sup>2</sup>,
</span>
<span class="author-block">
<a href="https://blender.cs.illinois.edu/hengji.html">Heng Ji</a><sup>1</sup>,
</span>
<span class="author-block">
<a href="https://www.huan-zhang.com/">Huan Zhang</a><sup>1</sup>,
</span>
<span class="author-block">
<a href="https://tongzhang-ml.org/">Tong Zhang</a><sup>1</sup>
</span>
</div>
<div class="is-size-5 publication-authors" style="margin-top:1em;">
<span class="author-block"><sup>1</sup>University of Illinois Urbana-Champaign,</span>
<span class="author-block"><sup>2</sup>Northwestern University,</span>
<span class="author-block"><sup>3</sup>University of Toronto,</span>
<span class="author-block"><sup>4</sup>Toyota Technological Institute at Chicago</span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block"><sup>*</sup>Equal contribution</span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block" style="color:rgb(192, 54, 54);"><b>ICML 2025 Oral</b></span>
</div>
<div class="is-size-5 publication-authors" style="margin-top:0.8em;">
<a href="https://foundation-models-meet-embodied-agents.github.io/cvpr2026/" target="_blank" style="color:#d35400; font-weight:bold; text-decoration:none; background:linear-gradient(135deg, #fff5eb 0%, #ffe4cc 100%); padding:0.4em 0.8em; border-radius:6px; border:1px solid #e67e22; display:inline-block;">
🔥 EmbodiedBench Challenge @ <b>CVPR 2026 Workshop on Foundation Models Meet Embodied Agents (FMEA)</b> is now open!
</a>
</div>
<!-- Links -->
<div class="column has-text-centered" style="margin-top:1.5em;">
<span class="link-block">
<a href="challenge.html" class="external-link button is-normal is-rounded is-dark" style="background: #5e7154; border-color: #5e7154;">
<span class="icon">
<i class="fas fa-trophy"></i>
</span>
<span>Challenge</span>
</a>
</span>
<span class="link-block">
<a href="http://arxiv.org/abs/2502.09560" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>Paper</span>
</a>
</span>
<span class="link-block">
<a href="https://github.com/EmbodiedBench/EmbodiedBench" class="external-link button is-normal is-rounded is-dark" role="button" target="_blank">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code</span>
</a>
</span>
<span class="link-block">
<a href="https://huggingface.co/EmbodiedBench"
class="external-link button is-normal is-rounded is-dark">
<span class="icon" style="display: flex; align-items: center; height: 100%;">
<p style="font-size:18px; margin: 0;">🤗</p>
</span>
<span>Dataset</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- Teaser Video -->
<!-- <section class="hero teaser">
<div class="container is-max-desktop">
<div class="hero-body" style="padding-top:1rem;">
<video id="teaser" autoplay muted loop playsinline controls width="100%">
<source src="https://github.com/embodied-agent-eval/embodied-agent-eval.github.io/raw/main/website/videos/eai-0-overview.mp4" type="video/mp4">
Your browser does not support the video tag.
</video>
</div>
</div>
</section> -->
<!-- Short videos (Goal Interpretation, etc.) -->
<section class="section" style="padding-top:0.5rem;">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column">
<!-- <h3 class="title is-5">Goal Interpretation</h3> -->
<video autoplay muted loop playsinline controls width="100%">
<source src="https://github.com/EmbodiedBench/embodiedbench.github.io/raw/main/website/videos/example_success_mani.mp4" type="video/mp4">
</video>
</div>
<div class="column">
<!-- <h3 class="title is-5">Subgoal Decomposition</h3> -->
<video autoplay muted loop playsinline controls width="100%">
<source src="https://github.com/EmbodiedBench/embodiedbench.github.io/raw/main/website/videos/example_success_nav.mp4" type="video/mp4">
</video>
</div>
</div>
<p style="text-align:center; margin-top:1em; color:rgb(99, 95, 95); margin-top:0.7rem; margin-bottom:1rem; line-height: 1.55; font-size: 1.2rem;">
<b><span style="color: rgb(53, 131, 53); font-size: 1.2rem;">Successful</span> Examples in EmbodiedBench powered by GPT-4o: EB-Manipulation (left) and EB-Navigation (right).</b>
</p>
</div>
</section>
<section class="section" style="padding-top:0rem;">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column">
<!-- <h3 class="title is-5">Goal Interpretation</h3> -->
<video autoplay muted loop playsinline controls width="100%">
<source src="https://github.com/EmbodiedBench/embodiedbench.github.io/raw/main/website/videos/example_failure_mani.mp4" type="video/mp4">
</video>
</div>
<div class="column">
<!-- <h3 class="title is-5">Subgoal Decomposition</h3> -->
<video autoplay muted loop playsinline controls width="100%">
<source src="https://github.com/EmbodiedBench/embodiedbench.github.io/raw/main/website/videos/example_failure_nav.mp4" type="video/mp4">
</video>
</div>
</div>
<p style="text-align:center; margin-top:1em; color:rgb(99, 95, 95); margin-top:0.7rem; margin-bottom:1rem; line-height: 1.55; font-size: 1.2rem;">
<b><span style="color: rgb(187, 75, 62); font-size: 1.2rem;">Failure</span> Examples in EmbodiedBench powered by GPT-4o: EB-Manipulation (left) and EB-Navigation (right).</b>
</p>
</div>
</section>
<!-- Single-Paragraph Abstract Section -->
<!-- <section class="section" id="abstract">
<div class="container" style="max-width: 85%; font-size: 1.0em; margin: 0 auto; line-height: 1.55;">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Introduction</h2>
<div class="content has-text-justified">
<p style="text-align:justify; color:rgb(31, 30, 30);">
Leveraging Multi-modal Large Language Models (MLLMs) to create <b>vision-driven embodied agents</b>
offers a promising avenue for tackling real-world tasks. While language-centric embodied agents have
garnered substantial attention, MLLM-based embodied agents remain underexplored due to the lack of
comprehensive evaluation frameworks. To bridge this gap, we introduce <b>EmbodiedBench</b>, an
extensive benchmark designed to evaluate vision-driven embodied agents. EmbodiedBench features:
(1) <b>a diverse set of 1,128 testing tasks across four environments</b>, ranging from high-level semantic
tasks (e.g., household) to low-level tasks involving atomic actions (e.g., navigation and manipulation);
and (2) <b>six meticulously curated subsets evaluating essential agent capabilities</b> like common sense
reasoning, complex instruction following, spatial awareness, visual perception, and long-term planning.
Through extensive experiments, we evaluate 13 leading proprietary and open-source MLLMs within
EmbodiedBench. Our findings reveal that: MLLMs excel at high-level tasks but struggle with low-level
manipulation, with the best model, GPT-4o, scoring only on average. EmbodiedBench provides a
standardized, multifaceted evaluation platform that not only highlights existing challenges
but also offers valuable insights for advancing MLLM-based embodied agents.
</p>
</div>
</div>
</div>
</div>
</section> -->
<!-- Overview Section -->
<section class="section" id="overview" style="padding-top:1.5rem;">
<div class="container" style="max-width: 80%; margin: 0 auto;">
<h2 class="title is-3" style="text-align:center; margin-bottom:1rem; font-size: 2.3em;">Overview</h2>
<br>
<div class="content has-text-justified" style="max-width: 95%; font-size: 1.2em; margin: 0 auto; line-height: 1.55;">
<p style="text-align:justify; color:rgb(31, 30, 30);">
We introduce <b>EmbodiedBench</b>, a comprehensive benchmark designed to evaluate <b>Multi-modal Large Language Models (MLLMs) as embodied agents</b>.
While existing benchmarks have primarily focused on Large Language Models (LLMs) and high-level tasks, EmbodiedBench takes a leap forward by offering a comprehensive,
fine-grained evaluation of MLLM-based agents across both <b>high-level and low-level tasks</b>, as well as <b>six critical agent capabilities</b>.
</p>
<p style="text-align:justify; color:rgb(31, 30, 30);">
EmbodiedBench is more than just a benchmark-it's a <b>multifaceted,
standardized evaluation platform</b> that not only uncovers the current challenges in embodied AI but also provides actionable insights to push the boundaries of MLLM-driven embodied intelligence.
</p>
</div>
<!-- Figure 1 -->
<div class="columns is-centered has-text-centered">
<div class="column is-eight-fifths">
<figure class="table-figure">
<img class="figure-image" src="website/img/framework_overview.png" alt="Figure 1. Overview of EMBODIEDBENCH." style="max-width: 95%; text-align:center; margin-bottom:1rem;"/>
<figcaption class="figure-caption" style="font-size: 1.1em;">
<b>Figure 1.</b> Overview of EmbodiedBench. Two key features of our benchmark: various action levels and capability-oriented evaluation.
</figcaption>
</figure>
<p style="text-align:justify; color:rgb(31, 30, 30); margin-top:1rem; margin-bottom:1rem; line-height: 1.55; max-width: 95%; margin: 0 auto; font-size: 1.2em;">
EmbodiedBench is designed with two key features that set it apart from existing benchmarks:
<b>1. Diverse tasks with hierarchical action levels.</b> Among the four environments, <b>EB-ALFRED</b> and <b>EB-Habitat</b> focus on high-level task decomposition and planning (e.g., <i>"put a book on the desk"</i> ),
while <b>EB-Navigation</b> and <b>EB-Manipulation</b> demand planning with low-level actions (e.g., <i>translational/rotational control</i> ) and require precise perception and spatial reasoning.
<b>2. Capability-oriented evaluation.</b> Unlike previous benchmarks that primarily emphasize overall accuracy or module-specific performance,
EmbodiedBench introduces a fine-grained evaluation framework that assesses six critical capabilities of embodied agents, including <b>basic task solving</b>, <b>commonsense reasoning</b>, <b>complex instruction understanding</b>, <b>spatial awareness</b>, <b>visual perception</b>, and <b>long-horizon planning</b>.
</p>
</div>
</div>
<!-- Figure 2 -->
<br>
<br>
<h2 class="title is-3" style="text-align:center; margin-bottom: 2.5rem; font-size: 2.3em;">Agent Pipeline</h2>
<div class="columns is-centered has-text-centered" style="margin-top:2em;">
<div class="column is-eight-fifths">
<figure class="table-figure">
<img class="figure-image" src="website/img/planner_pipeline.png" alt="Figure 2. Vision-driven agent pipeline used in EMBODIEDBENCH." style="max-width: 98%; margin: 0 auto; margin-bottom:1rem;" />
<figcaption class="figure-caption" style="font-size: 1.1em;">
<b>Figure 2.</b> Vision-driven agent pipeline used in EmbodiedBench.
</figcaption>
</figure>
<p style="text-align:justify; margin-top:1em; margin-bottom:1rem; line-height: 1.55; max-width: 95%; font-size: 1.2em; margin: 0 auto; color:rgb(31, 30, 30);">
To evaluate MLLMs as agents in EmbodiedBench, we design a unified embodied agent pipeline, illustrated in Figure 2.
This pipeline provides a robust framework for processing multimodal inputs, reasoning through interactions, and generating structured, executable plans composed of sequential actions.
</p>
</div>
</div>
</div>
</section>
<!-- Leaderboard and Findings Section -->
<section class="section" id="leaderboard-findings" style="padding-top:1.5rem;">
<div class="container" style="max-width: 85%">
<h2 class="title is-3" style="text-align:center; margin-bottom:3rem; font-size: 2.3em;">Leaderboard</h2>
<h3 class="title is-4" style="text-align:center; margin-bottom:1rem; font-size: 1.7em;">High-Level Tasks</h3>
<ul class="nav nav-tabs" id="myTab1" role="tablist">
<li class="nav-item" role="presentation">
<button class="nav-link active" id="eb-alfred-tab" data-bs-toggle="tab"
data-bs-target="#eb-alfred-content" type="button" role="tab"
aria-controls="eb-alfred-content" aria-selected="true" style="font-size: 1.1em;">EB-ALFRED</button>
</li>
<li class="nav-item" role="presentation">
<button class="nav-link" id="eb-habitat-tab" data-bs-toggle="tab"
data-bs-target="#eb-habitat-content" type="button" role="tab"
aria-controls="eb-habitat-content" aria-selected="false" style="font-size: 1.1em;">EB-Habitat</button>
</li>
</ul>
<div class="tab-content" id="myTabContent">
<div class="tab-pane fade show active" id="eb-alfred-content" role="tabpanel"
aria-labelledby="eb-alfred-tab">
<div id="eb-alfred-benchmark-main-table"></div>
</div>
<div class="tab-pane fade" id="eb-habitat-content" role="tabpanel"
aria-labelledby="eb-habitat-tab">
<div id="eb-habitat-benchmark-main-table"></div>
</div>
</div>
<br>
<br>
<h3 class="title is-4" style="text-align:center; margin-bottom:1rem; font-size: 1.7em;">Low-Level Tasks</h3>
<ul class="nav nav-tabs" id="myTab2" role="tablist">
<li class="nav-item" role="presentation">
<button class="nav-link active" id="main-results-tab" data-bs-toggle="tab"
data-bs-target="#benchmark-table-content" type="button" role="tab"
aria-controls="main-results-tab" aria-selected="true" style="font-size: 1.1em;">EB-Navigation</button>
</li>
<li class="nav-item" role="presentation">
<button class="nav-link" id="eurus-code-table-tab" data-bs-toggle="tab"
data-bs-target="#eurus-code-table-content" type="button" role="tab"
aria-controls="eurus-code-table-tab" aria-selected="false" style="font-size: 1.1em;">EB-Manipulation</button>
</li>
</ul>
<div class="tab-content" id="myTabContent">
<div class="tab-pane fade show active" id="benchmark-table-content" role="tabpanel"
aria-labelledby="benchmark-table-content">
<div id="eb-navigation-benchmark-main-table"></div>
</div>
<div class="tab-pane fade" id="eurus-code-table-content" role="tabpanel"
aria-labelledby="eurus-code-table-content">
<div id="eb-manipulation-benchmark-main-table"></div>
</div>
</div>
<br>
<section class="section" id="overview" style="padding-top:4rem;">
<div class="container" style="max-width: 95%; font-size: 1.0em; margin: 0 auto; line-height: 1.55;">
<h2 class="title is-3" style="text-align:center; margin-bottom: 2.8rem; font-size: 2.3em;">Error Analysis</h2>
<p style="text-align:justify; margin-top:1em; color:rgb(31, 30, 30); margin-top:1.7rem; margin-bottom:1rem; line-height: 1.55; font-size: 1.2em;" >
In this section, We conducted an error analysis on GPT-4o to identify potential failure modes.
For each environment, we sample 10 failure episodes from each subset, resulting in a total of 110 failed episodes to be analyzed.
We found three main types of errors: perception errors, reasoning errors, and planning errors.
Each error type corresponds to a specific stage in our agent pipeline.
For example, perception errors occur during the visual state description stage, reasoning errors arise in the reflection and reasoning stages, and planning errors occur during the language plan and executable plan generation stages.
</p>
<figure class="table-figure" style="margin-top:1em; max-width: 90%; margin: 0 auto;">
<img class="figure-image" src="website/img/error_analysis.png" alt="Figure 3. Error Analysis." />
<figcaption class="figure-caption" style="text-align: center; font-size: 1.1em;">
<b>Figure 3.</b> Error Analysis.
</figcaption>
</figure>
</div>
</section>
<!-- Case Study -->
<section class="section" id="case-study" style="padding-top:2rem;">
<div class="container is-max-desktop">
<h2 class="title is-3" style="text-align:center; margin-bottom:1rem; font-size: 2.3em;">Case Study</h2>
<br>
<!-- Success Carousel -->
<h3 class="title is-4" style="text-align:center; margin-bottom:-0.1rem; font-size: 1.7em;">Success Examples</h3>
<div id="results-carousel" class="carousel results-carousel" style="margin-bottom: 2rem;">
<div class="box mx-auto my-4" style="max-width: 90%;">
<div class="content has-text-centered" style="margin-bottom: 2rem;">
<img src="website/img/success_alfred_claude.png" alt="" width="75%"/>
<p class="image-subtitle" style="font-size: 1.1em; margin-top: 1rem;">
<b>Figure 4.</b> Planning example of Claude-3.5-Sonnet in EB-ALFRED.
</p>
</div>
</div>
<div class="box mx-auto my-4" style="max-width: 90%;">
<div class="content has-text-centered" style="margin-bottom: 2rem;">
<img src="website/img/success_habitat_internvl.png" alt="" width="75%"/>
<p class="image-subtitle" style="font-size: 1.1em; margin-top: 1rem;">
<b>Figure 5.</b> Planning example in EB-Habitat for InternVL2.5-78B.
</p>
</div>
</div>
<div class="box mx-auto my-4" style="max-width: 90%;">
<div class="content has-text-centered" style="margin-bottom: 2rem;">
<img src="website/img/success_nav_gpt4o.png" alt="" width="75%" />
<p class="image-subtitle" style="font-size: 1.1em; margin-top: 1rem;">
<b>Figure 6.</b> Planning example of GPT-4o in EB-Navigation.
</p>
</div>
</div>
<div class="box mx-auto my-4" style="max-width: 90%;">
<div class="content has-text-centered" style="margin-bottom: 2rem;">
<img src="website/img/success_mani_gemini.png" alt="" width="75%" />
<p class="image-subtitle" style="font-size: 1.1em; margin-top: 1rem;">
<b>Figure 7.</b> Planning example of Gemini-1.5-pro in EB-Manipulation.
</p>
</div>
</div>
</div>
<!-- Failure Carousel -->
<br>
<h3 class="title is-4" style="text-align:center; margin-top: 2rem; margin-bottom:-0.1rem; font-size: 1.7em;">Failure Examples</h3>
<div id="results-carousel" class="carousel results-carousel" data-autoplay="true" data-pagination="false">
<div class="box mx-auto my-4" style="max-width: 90%;">
<div class="content has-text-centered" style="margin-bottom: 2rem;">
<img src="website/img/planning_error_alfred.png" alt="" width="70%"/>
<p class="image-subtitle" style="font-size: 1.1em; margin-top: 2rem; max-width: 70%; margin: auto;">
<b>Figure 8.</b> Planning Error Example in EB-ALFRED: The agent was supposed to locate "Book_2" by the 7th action but instead continued
interacting with the first book.
</p>
</div>
</div>
<div class="box mx-auto my-4" style="max-width: 90%;">
<div class="content has-text-centered" style="margin-bottom: 2rem;">
<img src="website/img/perception_error_mani.png" alt="" width="70%" style="margin-bottom: 1rem;"/>
<p class="image-subtitle" style="font-size: 1.1em; max-width: 70%; margin: auto;">
<b>Figure 9.</b> Perception Error Example in EB-Manipulation: the agent erroneously observed the color of the object.
</p>
</div>
</div>
<div class="box mx-auto my-4" style="max-width: 90%;">
<div class="content has-text-centered" style="margin-bottom: 2rem;">
<img src="website/img/reasoning_error_nav.png" alt="" width="70%" style="margin-bottom: 1rem;"/>
<p class="image-subtitle" style="font-size: 1.1em; max-width: 70%; margin: auto;">
<b>Figure 10.</b> Reasoning Error Example in EB-Navigation: the agent recognized it was blocked by the countertop but failed to attempt
navigating around it.
</p>
</div>
</div>
</div>
</div>
</section>
<!-- BibTeX Section -->
<section class="section" id="BibTeX" style="padding-top:1.5rem;">
<div class="container is-max-desktop content" style="max-width: 95%; margin: 0 auto;">
<h2 class="title is-3" style="text-align:center; font-size: 2.3em;">BibTeX</h2>
<pre style="font-size: 1.0em;">
@article{yang2025embodiedbench,
title={EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents},
author={Yang, Rui and Chen, Hanyang and Zhang, Junyu and Zhao, Mark and Qian, Cheng and Wang, Kangrui and Wang, Qineng and Koripella, Teja Venkat and Movahedi, Marziyeh and Li, Manling and others},
journal={arXiv preprint arXiv:2502.09560},
year={2025}
}
</pre>
</div>
</section>
<section>
<div class="section" id="org-banners" style="display:flex; margin-bottom: -3rem; margin-left: 3rem; margin-right: 3rem;">
<a href="https://illinois.edu/" target="_blank" rel="external" style="display: block; margin: auto;">
<img class="center-block org-banner" src="website/img/uiuc.webp" style="width: 350px; height: auto; display: block; margin: auto;">
</a>
<a href="https://www.northwestern.edu/" target="blank" class="ext-link" style="display: block; margin: auto;">
<img class="center-block org-banner" src="website/img/northwestern.png" style="width: 390px; height: auto; display: block; margin: auto;">
</a>
</div>
</section>
<section>
<div class="section" id="org-banners" style="display:flex; margin-top: -3rem; margin-left: 3rem; margin-right: 3rem;">
<a href="https://www.utoronto.ca/" target="blank" class="ext-link" style="display: block; margin: auto;">
<img class="center-block org-banner" src="website/img/toronto.png" style="width: 330px; height: auto; display: block; margin: auto;">
</a>
<a href="https://www.utoronto.ca/" target="blank" class="ext-link" style="display: block; margin: auto;">
<img class="center-block org-banner" src="website/img/ttic.jpg" style="width: 350px; height: auto; display: block; margin: auto;">
</a>
</div>
</section>
<section class="section" id="contact" style="padding-top: 1rem;">
<div class="container" style="max-width: 900px;">
<h2 class="title is-3" style="text-align:center; margin-bottom: 2rem; font-size: 2.3em;">Contact</h2>
<div class="contact-grid">
<div class="contact-card contact-card--channels">
<h3 class="title is-5">Stay Connected</h3>
<p>
Join the EmbodiedBench Slack workspace for announcements and discussion, or reach out through GitHub Issues for benchmark and challenge questions.
</p>
<div class="contact-actions">
<a href="https://join.slack.com/t/embodiedbench-flx2674/shared_invite/zt-3vtr5ifw3-~o1DFd3JkOjNhtBJ~vie6w" class="button is-dark is-rounded" target="_blank" rel="noopener noreferrer">
<span class="icon"><i class="fab fa-slack"></i></span>
<span>Join Slack</span>
</a>
<a href="https://github.com/EmbodiedBench/EmbodiedBench/issues" class="button is-dark is-rounded" target="_blank" rel="noopener noreferrer">
<span class="icon"><i class="fab fa-github"></i></span>
<span>GitHub Issues</span>
</a>
</div>
</div>
<div class="contact-card contact-card--wechat">
<h3 class="title is-5">WeChat Group</h3>
<p>Scan the QR code below to join the EmbodiedBench WeChat group.</p>
<img src="website/img/wechat-group-qr.jpg" alt="EmbodiedBench WeChat group QR code" class="contact-qr">
<p class="contact-note">The QR code image may expire periodically. If it stops working, please use Slack or GitHub to contact the organizers.</p>
</div>
</div>
</div>
</section>
<footer class="footer">
<!-- <div class="content has-text-centered"> -->
<div class="columns is-centered">
<div class="column is-9">
<div class="content">
<p style="font-size: 1.15em;">
This website is adapted from
<a href="https://nerfies.github.io/" style="text-decoration: none;">Nerfies</a>, licensed under a
<a rel="license" href="http://creativecommons.org/licenses/by-sa/4.0/" style="text-decoration: none;">
Creative Commons Attribution-ShareAlike 4.0 International License
</a>.
</p>
</div>
</div>
</div>
<!-- </div> -->
</footer>
</body>
</html>