Repository navigation
Expand file tree
/
Copy pathindex.html
More file actions
813 lines (776 loc) · 46.2 KB
/
Copy pathindex.html
File metadata and controls
813 lines (776 loc) · 46.2 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
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>OdysseyArena: Benchmarking Large Language Models for Long-Horizon, Active and Inductive Interactions</title>
<meta name="description"
content="OdysseyArena evaluates LLM agents on environments whose rules are hidden: agents must discover the transition dynamics through their own actions, over hundreds of steps, before they can plan.">
<meta name="color-scheme" content="light dark">
<meta name="theme-color" content="#f4f7f6">
<link rel="icon" href="public/favicon.ico" sizes="any">
<link rel="icon" type="image/svg+xml" href="public/logo.svg">
<link rel="icon" type="image/png" sizes="32x32" href="public/favicon-32.png">
<link rel="icon" type="image/png" sizes="16x16" href="public/favicon-16.png">
<link rel="apple-touch-icon" href="public/apple-touch-icon.png">
<!-- Open Graph / Twitter. Absolute URLs are required by social scrapers.
The page is served at https://os-copilot.github.io/OdysseyArena-Home/;
update every absolute URL together if the host changes. -->
<meta property="og:type" content="website">
<meta property="og:url" content="https://os-copilot.github.io/OdysseyArena-Home/">
<meta property="og:site_name" content="OdysseyArena">
<meta property="og:title" content="OdysseyArena: Benchmarking LLMs for Long-Horizon, Active and Inductive Interactions">
<meta property="og:description"
content="Environments whose rules are hidden. Agents must discover the transition dynamics through their own actions before they can plan.">
<meta property="og:image" content="https://os-copilot.github.io/OdysseyArena-Home/public/social/og-image.png">
<meta property="og:image:width" content="1200">
<meta property="og:image:height" content="630">
<meta name="twitter:card" content="summary_large_image">
<meta name="twitter:site" content="@qiushi_sun">
<meta name="twitter:title" content="OdysseyArena: Benchmarking LLMs for Long-Horizon, Active and Inductive Interactions">
<meta name="twitter:description"
content="Environments whose rules are hidden. Agents must discover the transition dynamics through their own actions before they can plan.">
<meta name="twitter:image" content="https://os-copilot.github.io/OdysseyArena-Home/public/social/og-image.png">
<link rel="preconnect" href="https://fonts.googleapis.com">
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link rel="stylesheet"
href="https://fonts.googleapis.com/css2?family=Bricolage+Grotesque:opsz,wght@12..96,500;12..96,600;12..96,700&family=Inter:wght@400;500;600;700&family=IBM+Plex+Mono:wght@400;500;600&display=swap">
<link rel="stylesheet" href="static/css/main.css">
<link rel="stylesheet" href="static/css/hero.css">
<!-- Reveal-on-scroll hides content until JS un-hides it, so only opt in when
JS is running; and the theme must be set before first paint. Light is the
default; a stored choice wins. -->
<script>
(function () {
var d = document.documentElement;
d.className += " js";
var t = (location.search.match(/[?&]theme=(light|dark)/) || [])[1];
if (!t) { try { t = localStorage.getItem("odyssey-theme"); } catch (e) { /* private mode */ } }
d.setAttribute("data-theme", t === "dark" ? "dark" : "light");
})();
</script>
</head>
<body>
<a class="skip-link" href="#main">Skip to content</a>
<div class="cosmos" aria-hidden="true"></div>
<!-- ==================================================================== -->
<!-- Navigation -->
<!-- ==================================================================== -->
<nav class="nav" id="nav">
<div class="wrap">
<a class="nav-brand" href="#top">
<img src="public/logo.svg" alt="">
<span>OdysseyArena</span>
</a>
<button class="icon-btn nav-toggle" id="navToggle" aria-expanded="false" aria-controls="navLinks" aria-label="Menu">
<svg aria-hidden="true" width="17" height="17" viewBox="0 0 16 16" fill="none" stroke="currentColor" stroke-width="1.6">
<path d="M2 4h12M2 8h12M2 12h12" stroke-linecap="round" />
</svg>
</button>
<ul class="nav-links" id="navLinks">
<li class="nav-more" id="navMore" data-open="false">
<button type="button" aria-expanded="false" aria-haspopup="true">
More research
<svg viewBox="0 0 10 10" fill="none" stroke="currentColor" stroke-width="1.6" aria-hidden="true">
<path d="M1.5 3.5 5 7l3.5-3.5" stroke-linecap="round" stroke-linejoin="round" />
</svg>
</button>
<ul>
<li><a href="https://os-copilot.github.io/OSReward-Home/" target="_blank" rel="noopener">
<img src="public/logos/projects/osreward.png" alt="">
<span>OSReward <small>CUA Reward Eval & Open Models</small></span></a></li>
<li><a href="https://os-copilot.github.io/OpenMobile2-Home/" target="_blank" rel="noopener">
<img src="public/logos/projects/openmobile.png" alt="">
<span>OpenMobile-2 <small>Versatile Mobile Agents</small></span></a></li>
<li><a href="https://qiushisun.github.io/OS-Sentinel-Home/" target="_blank" rel="noopener">
<img src="public/logos/projects/os-sentinel.png" alt="">
<span>OS-Sentinel <small>Mobile GUI Safety</small></span></a></li>
<li><a href="https://qiushisun.github.io/ScienceBoard-Home/" target="_blank" rel="noopener">
<img src="public/logos/projects/scienceboard.png" alt="">
<span>ScienceBoard <small>CUA for Science</small></span></a></li>
<li><a href="https://qiushisun.github.io/OS-Genesis-Home/" target="_blank" rel="noopener">
<img src="public/logos/projects/os-genesis.png" alt="">
<span>OS-Genesis <small>Annotation-Free Trajectory Synthesis</small></span></a></li>
<li><a href="https://osatlas.github.io/" target="_blank" rel="noopener">
<img src="public/logos/projects/os-atlas.png" alt="">
<span>OS-Atlas <small>GUI Action Foundation Model</small></span></a></li>
<li><a href="https://njucckevin.github.io/openmobile/" target="_blank" rel="noopener">
<img src="public/logos/projects/openmobile.png" alt="">
<span>OpenMobile <small>Open Mobile Agents</small></span></a></li>
<li><a href="https://aclanthology.org/2024.acl-long.505" target="_blank" rel="noopener">
<img src="public/logos/projects/seeclick.png" alt="">
<span>SeeClick <small>GUI Grounding</small></span></a></li>
</ul>
</li>
<!-- Section links double as the scroll-spy list (main.js reads them). -->
<li><a href="#overview">Overview</a></li>
<li><a href="#environments">Environments</a></li>
<li><a href="#benchmark">Benchmark</a></li>
<li><a href="#results">Results</a></li>
<li><a href="#analysis">Analysis</a></li>
<li><a href="#resources">Resources</a></li>
</ul>
<div class="nav-tools">
<button class="icon-btn" id="themeToggle" type="button" aria-label="Switch to dark theme">
<svg class="i-moon" viewBox="0 0 16 16" fill="none" stroke="currentColor" stroke-width="1.4" aria-hidden="true">
<path d="M14 9.3A6 6 0 0 1 6.7 2a6.2 6.2 0 1 0 7.3 7.3Z" stroke-linejoin="round" />
</svg>
<svg class="i-sun" viewBox="0 0 16 16" fill="none" stroke="currentColor" stroke-width="1.4" aria-hidden="true">
<circle cx="8" cy="8" r="3.1" />
<path d="M8 1v1.7M8 13.3V15M1 8h1.7M13.3 8H15M3 3l1.2 1.2M11.8 11.8 13 13M13 3l-1.2 1.2M4.2 11.8 3 13" stroke-linecap="round" />
</svg>
</button>
<a class="icon-btn" href="https://github.com/xufangzhi/Odyssey-Arena" target="_blank" rel="noopener" aria-label="GitHub">
<svg viewBox="0 0 16 16" fill="currentColor" aria-hidden="true"><path d="M8 0C3.58 0 0 3.58 0 8a8 8 0 0 0 5.47 7.59c.4.07.55-.17.55-.38 0-.19-.01-.82-.01-1.49-2.01.37-2.53-.49-2.69-.94-.09-.23-.48-.94-.82-1.13-.28-.15-.68-.52-.01-.53.63-.01 1.08.58 1.23.82.72 1.21 1.87.87 2.33.66.07-.52.28-.87.51-1.07-1.78-.2-3.64-.89-3.64-3.95 0-.87.31-1.59.82-2.15-.08-.2-.36-1.02.08-2.12 0 0 .67-.21 2.2.82a7.4 7.4 0 0 1 2-.27c.68 0 1.36.09 2 .27 1.53-1.04 2.2-.82 2.2-.82.44 1.1.16 1.92.08 2.12.51.56.82 1.27.82 2.15 0 3.07-1.87 3.75-3.65 3.95.29.25.54.73.54 1.48 0 1.07-.01 1.93-.01 2.2 0 .21.15.46.55.38A8.01 8.01 0 0 0 16 8c0-4.42-3.58-8-8-8Z" /></svg>
</a>
</div>
</div>
</nav>
<!-- ==================================================================== -->
<!-- Hero: content left, the playable environment right -->
<!-- ==================================================================== -->
<header class="hero" id="top">
<div class="wrap wrap-wide hero-split">
<div class="hero-content">
<h1><span class="grad">OdysseyArena</span></h1>
<p class="tagline">Benchmarking Large Language Models for Long-Horizon, Active and Inductive Interactions</p>
<!-- TODO: venue line on acceptance. <p class="venue">…</p> -->
<p class="lede">Most agent benchmarks hand the agent the rules and score how well it follows them.
OdysseyArena hides them. Its environments evolve under latent transition laws that an agent must
discover through its own actions, over hundreds of steps, before it can plan.</p>
<ul class="bullets">
<li>
<span class="b-ico" aria-hidden="true">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.8" stroke-linecap="round" stroke-linejoin="round">
<path d="M9 18h6M10 21h4M12 3a6 6 0 0 0-3.6 10.8c.6.5.6 1.2.6 1.7V16h6v-.5c0-.5 0-1.2.6-1.7A6 6 0 0 0 12 3Z" />
</svg>
</span>
<span><b>Inductive evaluation</b>: four structural primitives, from Boolean couplings to dependency
graphs, each turned into an environment whose rules are never shown to the agent</span>
</li>
<li>
<span class="b-ico" aria-hidden="true">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.8" stroke-linecap="round" stroke-linejoin="round">
<path d="M3 12h4l3-8 4 16 3-8h4" />
</svg>
</span>
<span><b>OdysseyArena-Lite and OdysseyArena-Challenge</b>: a standardized suite for routine evaluation,
and a stress test whose trajectories run past a thousand steps</span>
</li>
<li>
<span class="b-ico" aria-hidden="true">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.8" stroke-linecap="round">
<path d="M4 19V9M10 19V5M16 19v-6M22 19H2" />
</svg>
</span>
<span><b>Findings on frontier LLMs</b>: near-perfect once the rules are given, far below humans without
them, with the failures traced to exploration, memory and action loops rather than planning</span>
</li>
</ul>
<!-- Release links. Paper on arXiv; code and the task files on GitHub. -->
<div class="links">
<a class="btn btn-primary" data-link="paper" href="https://arxiv.org/abs/2602.05843" target="_blank" rel="noopener">
<svg viewBox="0 0 17.732 24.269" aria-hidden="true">
<g transform="translate(-566.984 -271.548)">
<path d="M573.549,280.916l2.266,2.738,6.674-7.84c.353-.47.52-.717.353-1.117a1.218,1.218,0,0,0-1.061-.748h0a.953.953,0,0,0-.712.262Z" fill="currentColor" />
<path d="M579.525,282.225l-10.606-10.174a1.413,1.413,0,0,0-.834-.5,1.09,1.09,0,0,0-1.027.66c-.167.4-.047.681.319,1.206l8.44,10.242h0l-6.282,7.716a1.336,1.336,0,0,0-.323,1.3,1.114,1.114,0,0,0,1.04.69A.992.992,0,0,0,571,293l8.519-7.92A1.924,1.924,0,0,0,579.525,282.225Z" fill="#b31b1b" />
<path d="M584.32,293.912l-8.525-10.275,0,0L573.53,280.9l-1.389,1.254a2.063,2.063,0,0,0,0,2.965l10.812,10.419a.925.925,0,0,0,.742.282,1.039,1.039,0,0,0,.953-.667A1.261,1.261,0,0,0,584.32,293.912Z" fill="currentColor" />
</g>
</svg>
Paper
</a>
<a class="btn" data-link="code" href="https://github.com/xufangzhi/Odyssey-Arena" target="_blank" rel="noopener">
<svg viewBox="0 0 16 16" fill="currentColor" aria-hidden="true">
<path d="M8 0C3.58 0 0 3.58 0 8a8 8 0 0 0 5.47 7.59c.4.07.55-.17.55-.38 0-.19-.01-.82-.01-1.49-2.01.37-2.53-.49-2.69-.94-.09-.23-.48-.94-.82-1.13-.28-.15-.68-.52-.01-.53.63-.01 1.08.58 1.23.82.72 1.21 1.87.87 2.33.66.07-.52.28-.87.51-1.07-1.78-.2-3.64-.89-3.64-3.95 0-.87.31-1.59.82-2.15-.08-.2-.36-1.02.08-2.12 0 0 .67-.21 2.2.82a7.4 7.4 0 0 1 2-.27c.68 0 1.36.09 2 .27 1.53-1.04 2.2-.82 2.2-.82.44 1.1.16 1.92.08 2.12.51.56.82 1.27.82 2.15 0 3.07-1.87 3.75-3.65 3.95.29.25.54.73.54 1.48 0 1.07-.01 1.93-.01 2.2 0 .21.15.46.55.38A8.01 8.01 0 0 0 16 8c0-4.42-3.58-8-8-8Z" />
</svg>
Code
</a>
<a class="btn" data-link="benchmark" href="https://github.com/xufangzhi/Odyssey-Arena/tree/main/test_data" target="_blank" rel="noopener">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.8" aria-hidden="true">
<ellipse cx="12" cy="6" rx="8" ry="3" />
<path d="M4 6v6c0 1.66 3.58 3 8 3s8-1.34 8-3V6M4 12v6c0 1.66 3.58 3 8 3s8-1.34 8-3v-6" />
</svg>
Benchmark
</a>
<a class="btn" href="#results">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.8" aria-hidden="true">
<path d="M8 21h8M12 17v4M5 3h14v5a7 7 0 0 1-14 0V3Z" stroke-linejoin="round" />
<path d="M5 5H3v2a3 3 0 0 0 3 3M19 5h2v2a3 3 0 0 1-3 3" stroke-linejoin="round" />
</svg>
Leaderboard
</a>
</div>
</div>
<!-- Key art: the four environments, live. hero.js runs a scripted agent on
a fresh instance of each; a click opens that environment below. -->
<figure class="hero-art">
<div id="arena" aria-label="The four environments, each played live by a scripted agent"></div>
<figcaption>Four environments, live. Hover to pause; click one to step through a real trajectory or play it yourself.</figcaption>
</figure>
</div>
</header>
<!-- Byline. Equal contribution is marked by colour (the .eq authors).
TODO: affiliations once the list is settled (crests as 96x96 PNG pairs
in public/affil/, listed in an .affils row under the names). -->
<section class="byline" aria-label="Authors">
<div class="panel">
<p class="authors">
<span class="lead">
<span class="au eq">Fangzhi Xu</span>,
<span class="au eq">Hang Yan</span>,
<span class="au eq"><a href="https://qiushisun.github.io/" target="_blank" rel="noopener">Qiushi Sun</a><sup class="author-mark" aria-label="Corresponding author" title="Corresponding author">✉︎</sup></span>
</span>
<span class="rest">
<span class="au">Jinyang Wu</span>,
<span class="au">Zixian Huang</span>,
<span class="au">Muye Huang</span>,
<span class="au">Jingyang Gong</span>,
<span class="au">Zichen Ding</span>,
<span class="au">Kanzhi Cheng</span>,
<span class="au">Yian Wang</span>,
<span class="au">Xinyu Che</span>,
<span class="au">Zeyi Sun</span>,
<span class="au">Jian Zhang</span>,
<span class="au">Zhangyue Yin</span>,
<span class="au">Haoran Luo</span>,
<span class="au">Ben Kao</span>,
<span class="au">Qika Lin<sup class="author-mark" aria-label="Corresponding author" title="Corresponding author">✉︎</sup></span>
</span>
</p>
<p class="author-legend">
<span class="eq-key">Equal Contribution</span>
<span class="author-role"><span class="role-mark" aria-hidden="true">✉︎</span> Corresponding authors</span>
</p>
</div>
</section>
<section class="news-band" aria-label="News">
<div class="wrap">
<div class="card news rise">
<span class="eyebrow">News</span>
<ul>
<li><time datetime="2026-02">Feb 2026</time>The paper is on <a href="https://arxiv.org/abs/2602.05843" target="_blank" rel="noopener">arXiv</a>.</li>
<!-- TODO: venue, further releases. -->
</ul>
</div>
</div>
</section>
<main id="main">
<!-- ================================================================== -->
<!-- Overview -->
<!-- ================================================================== -->
<section class="band-tight" id="overview">
<div class="wrap">
<div class="section-head">
<span class="eyebrow">Overview</span>
<h2 class="section-title">Agents that must <span class="accent">discover the rules</span></h2>
<p class="section-sub">Interactive benchmarks so far test a deductive mode: the agent is told how the
world works and is scored on carrying out a task in it. Deployment also needs the inductive mode,
where the agent works out how the world works from what its actions do.</p>
</div>
<div class="overview-grid rise">
<div class="card dg-card">
<div class="dg">
<div class="dg-pane ded">
<h4><i></i>Deductive</h4>
<svg viewBox="0 0 300 150" aria-label="Deductive setting: rules are given to the agent, which acts and observes">
<defs>
<marker id="arD" viewBox="0 0 10 10" refX="8" refY="5" markerWidth="6" markerHeight="6" orient="auto">
<path d="M0 0L10 5 0 10z" fill="var(--blue-400)" />
</marker>
</defs>
<g transform="translate(16,38)">
<rect class="rules-doc" x="0" y="0" width="54" height="66" rx="6" />
<line class="rules-line" x1="11" y1="17" x2="43" y2="17" />
<line class="rules-line" x1="11" y1="29" x2="39" y2="29" />
<line class="rules-line" x1="11" y1="41" x2="43" y2="41" />
<line class="rules-line" x1="11" y1="53" x2="33" y2="53" />
</g>
<text class="lbl k" x="43" y="124" text-anchor="middle">rules</text>
<path d="M76 71H114" fill="none" stroke="var(--blue-400)" stroke-width="1.8" marker-end="url(#arD)" />
<g transform="translate(120,48)">
<line x1="22" y1="2" x2="22" y2="9" stroke="var(--text-dim)" stroke-width="1.5" />
<circle cx="22" cy="0" r="2.4" class="bot-eye" />
<rect class="bot" x="2" y="9" width="40" height="34" rx="9" />
<circle class="bot-eye" cx="15" cy="25" r="3.4" />
<circle class="bot-eye" cx="29" cy="25" r="3.4" />
</g>
<text class="lbl k" x="142" y="124" text-anchor="middle">agent</text>
<g transform="translate(246,71)">
<circle class="world" r="27" />
<ellipse class="world-h" rx="11" ry="27" />
<ellipse class="world-h" rx="27" ry="10" />
</g>
<text class="lbl k" x="246" y="124" text-anchor="middle">environment</text>
<path class="arc" d="M168 56C190 34 206 34 222 48" marker-end="url(#arD)" />
<text class="lbl" x="196" y="32" text-anchor="middle">act</text>
<path class="arc" d="M220 96C206 110 190 110 168 88" marker-end="url(#arD)" />
<text class="lbl" x="196" y="119" text-anchor="middle">observe</text>
</svg>
<p>The transition rules arrive in the prompt. Success measures how well the agent applies them.</p>
</div>
<div class="dg-pane ind">
<h4><i></i>Inductive</h4>
<svg viewBox="0 0 300 150" aria-label="Inductive setting: the agent probes an unknown environment and induces its rules">
<defs>
<marker id="arI" viewBox="0 0 10 10" refX="8" refY="5" markerWidth="6" markerHeight="6" orient="auto">
<path d="M0 0L10 5 0 10z" fill="var(--amber-500)" />
</marker>
</defs>
<g transform="translate(14,10)">
<rect class="thought" x="0" y="0" width="76" height="54" rx="10" />
<g>
<line class="thought-line" x1="12" y1="16" x2="62" y2="16" />
<line class="thought-line" x1="12" y1="28" x2="52" y2="28" />
<line class="thought-line" x1="12" y1="40" x2="58" y2="40" />
</g>
<circle class="thought" cx="84" cy="62" r="4" />
<circle class="thought" cx="94" cy="72" r="2.4" />
</g>
<text class="lbl k" x="52" y="80" text-anchor="middle">induced rules</text>
<g transform="translate(96,56)">
<line x1="22" y1="2" x2="22" y2="9" stroke="var(--text-dim)" stroke-width="1.5" />
<circle cx="22" cy="0" r="2.4" class="bot-eye" />
<rect class="bot" x="2" y="9" width="40" height="34" rx="9" />
<circle class="bot-eye" cx="15" cy="25" r="3.4" />
<circle class="bot-eye" cx="29" cy="25" r="3.4" />
</g>
<text class="lbl k" x="118" y="124" text-anchor="middle">agent</text>
<g transform="translate(240,71)">
<circle class="world" r="27" />
<ellipse class="world-h" rx="11" ry="27" />
<ellipse class="world-h" rx="27" ry="10" />
<circle r="11" fill="var(--surface-3)" />
<text class="qmark" x="0" y="6" text-anchor="middle">?</text>
</g>
<text class="lbl k" x="240" y="124" text-anchor="middle">environment</text>
<path class="arc" d="M144 60C172 34 198 34 216 48" marker-end="url(#arI)" />
<text class="lbl" x="180" y="32" text-anchor="middle">act, probe</text>
<path class="arc" d="M214 96C198 110 172 110 144 92" marker-end="url(#arI)" />
<text class="lbl" x="180" y="119" text-anchor="middle">observe, induce</text>
</svg>
<p>Nothing about the dynamics is given. The agent has to probe, read the feedback and build the
rules itself.</p>
</div>
</div>
</div>
<div class="dims">
<div class="cap">
<span class="ico" aria-hidden="true">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.7" stroke-linecap="round" stroke-linejoin="round">
<path d="M3 17c3-6 5-6 8 0s5 6 8 0" stroke-dasharray="3 2.5" />
<circle cx="21" cy="12" r="2.2" fill="currentColor" stroke="none" />
</svg>
</span>
<div>
<h3>Extremely long horizons</h3>
<p>Episodes of hundreds to thousands of steps, where strategic coherence and accumulated error
decide the outcome, where most suites stop under fifty.</p>
</div>
</div>
<div class="cap">
<span class="ico" aria-hidden="true">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.7" stroke-linecap="round" stroke-linejoin="round">
<circle cx="12" cy="12" r="9" />
<path d="M12 3v3M12 18v3M3 12h3M18 12h3" />
<path d="m15 9-2 5-4 1 1-4 5-2Z" fill="currentColor" stroke="none" opacity="0.85" />
</svg>
</span>
<div>
<h3>Active exploration</h3>
<p>No fully specified success criterion to execute against. The agent probes, reacts to feedback
and revises its own strategy as it goes.</p>
</div>
</div>
<div class="cap">
<span class="ico" aria-hidden="true">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.7" stroke-linecap="round" stroke-linejoin="round">
<path d="M4 20h16" />
<path d="M6 20V9l6-4 6 4v11" />
<path d="M12 20v-5" />
<path d="M9 12h6" stroke-dasharray="2 2" />
</svg>
</span>
<div>
<h3>Inductive reasoning from interaction</h3>
<p>Latent transition laws are inferred from real-time experience, which is what lets an agent
anticipate outcomes and plan instead of react.</p>
</div>
</div>
</div>
</div>
<div class="card contrib-card rise" style="margin-top:22px">
<ol class="contribs">
<li><a href="#environments"><b>A new perspective for agent evaluation.</b></a> Agents are scored on
what they can discover about an environment's dynamics through long-horizon, active interaction,
rather than on following rules they are handed.</li>
<li><a href="#benchmark"><b>Reliable and scalable evaluation.</b></a> Four environments built on four
structural primitives, each with an unbounded task space; OdysseyArena-Lite fixes a standard suite
for efficient, reproducible comparison and OdysseyArena-Challenge stress-tests stability over
extreme horizons.</li>
<li><a href="#results"><b>Extensive evaluation and insights.</b></a> Frontier and open-weight LLMs,
with an analysis that locates the bottleneck in inductive discovery: exploration, memory and action
loops, not planning.</li>
</ol>
</div>
</div>
</section>
<!-- ================================================================== -->
<!-- Environments -->
<!-- ================================================================== -->
<section class="band" id="environments">
<div class="wrap">
<div class="section-head">
<span class="eyebrow">Environments</span>
<h2 class="section-title">Four primitives, <span class="accent">four environments</span></h2>
<p class="section-sub">Every environment is a transition function the agent cannot see. Each one
isolates a different kind of structure to be discovered: Boolean couplings, factors under noise,
periodic cycles, and a dependency graph. All four run as lightweight APIs, with no simulator or
container to deploy.</p>
</div>
<div class="card ew-card rise" id="env-walk"></div>
<p class="wt-note">Trajectories follow the demonstrations in the paper; the prices, news and package names
are schematic. Use the arrow keys or Play to step through.</p>
</div>
</section>
<!-- ================================================================== -->
<!-- Benchmark -->
<!-- ================================================================== -->
<section class="band" id="benchmark">
<div class="wrap">
<div class="section-head">
<span class="eyebrow">Benchmark</span>
<h2 class="section-title">Two suites from <span class="accent">one task space</span></h2>
<p class="section-sub">Each environment generates tasks from a bounded parameter range. A task is a
structural configuration that stays fixed for the episode plus a pre-computed temporal trajectory,
so every model faces exactly the same world and every instance is verified solvable.</p>
</div>
<div class="suites rise">
<div class="card suite lite">
<span class="tag"><i></i>Standard evaluation</span>
<h3>OdysseyArena-Lite</h3>
<p>The suite behind every result on this page: a tractable parameter range, run at high throughput,
with horizons long enough that nothing short of rule induction gets through.</p>
<dl>
<dt>Tasks</dt><dd id="lite-tasks"></dd>
<dt>Step budget</dt><dd id="lite-steps"></dd>
<dt>Protocol</dt><dd id="lite-runs"></dd>
</dl>
<div class="horizon"><span class="lab">0</span><div class="track"><i id="lite-horizon"></i></div><span class="lab">200 steps</span></div>
</div>
<div class="card suite chal">
<span class="tag"><i></i>Stress test</span>
<h3>OdysseyArena-Challenge</h3>
<p>The empirical limits of the same parameter ranges, for failure modes that never surface under a
standard budget: credit assignment over very long histories and compounding planning errors.</p>
<dl>
<dt>Tasks</dt><dd id="chal-tasks"></dd>
<dt>Horizon</dt><dd id="chal-steps"></dd>
<dt>Use</dt><dd>Optional; not part of the main evaluation</dd>
</dl>
<div class="horizon"><span class="lab">0</span><div class="track"><i id="chal-horizon"></i></div><span class="lab">1,000+ steps</span></div>
</div>
</div>
<div class="card bf-card rise" style="margin-top:22px">
<div class="chart-head"><h3>From a configuration to a verdict</h3></div>
<p class="chart-note">A task is written once as a configuration, generated and validated automatically,
then run through one interaction loop: the agent reads the observation and its working memory,
emits an action, and the environment's step logic answers.</p>
<div class="bf">
<div class="bf-col">
<div class="bf-node cfg">
<h4><span class="k">1</span>Configuration</h4>
<p>Entity counts, dependency depth, noise, periods and thresholds, set by hand per difficulty.</p>
<span class="ok"><svg viewBox="0 0 16 16" fill="none" stroke="currentColor" stroke-width="1.8"><path d="m3 8 3 3 7-7" stroke-linecap="round" stroke-linejoin="round" /></svg>difficulty control</span>
</div>
<div class="bf-node">
<h4><span class="k">2</span>Task generator</h4>
<p>Samples the latent rules and the fixed temporal trajectory, then checks a solution exists.</p>
<span class="ok"><svg viewBox="0 0 16 16" fill="none" stroke="currentColor" stroke-width="1.8"><path d="m3 8 3 3 7-7" stroke-linecap="round" stroke-linejoin="round" /></svg>automatic validation</span>
</div>
<div class="bf-node">
<h4><span class="k">3</span>Config files</h4>
<p>One deterministic instance per file; identical for every model and every run.</p>
</div>
</div>
<div class="bf-arrow" aria-hidden="true">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.8" stroke-linecap="round" stroke-linejoin="round"><path d="M5 12h14M13 6l6 6-6 6" /></svg>
</div>
<div class="bf-loop" id="bf-loop">
<div class="bf-agent">
<div>LLM agent<small>thought + action</small></div>
<div>Working memory<small>actions and feedback so far</small></div>
</div>
<div class="bf-wire" aria-hidden="true">
<svg viewBox="0 0 420 34">
<path d="M60 2v10q0 8 8 8h142" />
<path d="M360 2v10q0 8-8 8H210" />
<text x="210" y="31" text-anchor="middle">observation in, action out</text>
</svg>
</div>
<div class="ring">
<div class="bf-fn"><code>init()</code><small>load the instance</small></div>
<div class="bf-fn"><code>step()</code><small>apply the action, return the observation</small></div>
<div class="bf-fn"><code>eval()</code><small>check the terminal state</small></div>
</div>
<div class="bf-out">
<span class="win"><i></i>task success</span>
<span class="fail"><i></i>task failure</span>
</div>
</div>
</div>
</div>
<div class="card rise" style="margin-top:22px">
<div class="chart-head"><h3>Against other agent benchmarks</h3></div>
<p class="chart-note">Whether a benchmark requires inducing rules, how many steps a task takes (short
under fifty, long up to a hundred, extra long beyond), and what has to be deployed to run it.</p>
<div class="table-scroll">
<table class="cmp">
<thead><tr><th>Benchmark</th><th>Inductive</th><th>Horizon</th><th>Deploy</th></tr></thead>
<tbody id="cmp-body"></tbody>
</table>
</div>
</div>
</div>
</section>
<!-- ================================================================== -->
<!-- Results -->
<!-- ================================================================== -->
<section class="band" id="results">
<div class="wrap wrap-wide">
<div class="section-head">
<span class="eyebrow">Results</span>
<h2 class="section-title">How far <span class="accent">frontier models</span> get</h2>
<p class="section-sub">Proprietary and open-weight LLMs on OdysseyArena-Lite, four runs per task, against
human participants who solved the same instances by trial and error.</p>
</div>
<div class="findings rise">
<div class="finding">
<span class="k">A</span>
<div><b>A wide gap to humans</b>
<p>Human participants outperform every evaluated model on Avg@4 in Turn On Lights, AI Trading
and Repo System. Gemini 3 Pro Preview leads on Energy Dispatch.</p></div>
</div>
<div class="finding">
<span class="k">B</span>
<div><b>Deductive proficiency, inductive deficiency</b>
<p>Given the latent rules in the prompt, frontier models reach near-perfect success on Turn On
Lights. Without them the same models fall to a fraction. The bottleneck is discovering the
dynamics, not executing them.</p></div>
</div>
<div class="finding">
<span class="k">C</span>
<div><b>Periodic patterns defeat almost everyone</b>
<p>Energy Dispatch needs about twenty steps of observation to pin down an efficiency cycle. Nearly
every model fails every task there, and scale alone does not change that.</p></div>
</div>
</div>
<div class="card chart-card rise">
<div class="chart-head"><h3>Leaderboard</h3></div>
<p class="chart-note">Success rate per environment; profit rate for AI Trading. Click a column to sort.</p>
<div id="leaderboard"></div>
</div>
<div class="card chart-card rise" style="margin-top:22px">
<div class="chart-head"><h3>The same models with the rules given</h3></div>
<p class="chart-note">Each model run twice on the same tasks: as in the benchmark, and with the hidden
transition rules written into the prompt.</p>
<div class="controls"><span class="ctl-label">Environment</span><div class="seg" id="rules-env" role="group" aria-label="Environment"></div></div>
<div id="chart-rules"></div>
</div>
</div>
</section>
<!-- ================================================================== -->
<!-- Analysis -->
<!-- ================================================================== -->
<section class="band" id="analysis">
<div class="wrap">
<div class="section-head">
<span class="eyebrow">Analysis</span>
<h2 class="section-title">Where the <span class="accent">runs go wrong</span></h2>
<p class="section-sub">Failed trajectories were read by hand and the behaviour measured over every run:
what a model does with more steps, how often it repeats a rejected action, and which interventions
move the number.</p>
</div>
<div class="grid grid-2 rise">
<div class="card chart-card">
<div class="chart-head"><h3>More steps stop helping</h3></div>
<p class="chart-note">Cumulative share of tasks solved by each step, or profit rate for AI Trading. Most success curves flatten after an early
exploratory phase; a longer budget does not substitute for a model of the dynamics. The weakest
models track or trail a random policy.</p>
<div class="controls"><div class="seg" id="curves-env" role="group" aria-label="Environment"></div></div>
<div id="chart-curves"></div>
</div>
<div class="card chart-card">
<div class="chart-head"><h3>Action loops</h3></div>
<p class="chart-note">The share of a model's steps spent immediately repeating an action the
environment just rejected, against its success rate. Looping is inductive stagnation made
visible: the model does not turn the failed trial into a revised rule.</p>
<div class="controls"><div class="seg" id="loops-env" role="group" aria-label="Environment"></div></div>
<div id="chart-loops"></div>
</div>
</div>
<div class="card chart-card rise" style="margin-top:22px">
<div class="chart-head"><h3>What went wrong in the failed runs</h3></div>
<p class="chart-note">Every failed trajectory of Gemini 3 Pro Preview, read and labelled by hand; one
trajectory can carry several modes. Planning is almost never the problem.</p>
<div id="chart-failures"></div>
</div>
<div class="mini-grid rise" style="margin-top:22px">
<div class="card mini">
<h3>More reasoning helps</h3>
<p class="sub">gpt-oss-120b at three reasoning budgets. Higher effort improves Turn On Lights and AI Trading.
Repo System dips at medium effort; Energy Dispatch stays at zero.</p>
<div id="chart-effort"></div>
</div>
<div class="card mini">
<h3>A longer memory does not</h3>
<p class="sub">Repo System with more or fewer turns of history in the prompt. A window of about a hundred turns does best;
the full history does worse.</p>
<div id="chart-memory"></div>
</div>
<div class="card mini">
<h3>A worked example does not either</h3>
<p class="sub">One complete solution in the prompt leaves Repo System success where it was.
Eliciting induction takes more than a demonstration.</p>
<div id="chart-icl"></div>
</div>
</div>
</div>
</section>
<!-- ================================================================== -->
<!-- Resources -->
<!-- ================================================================== -->
<section class="band" id="resources">
<div class="wrap">
<div class="section-head">
<span class="eyebrow">Resources</span>
<h2 class="section-title">Paper, code and <span class="accent">benchmark</span></h2>
</div>
<div class="grid grid-4 rise">
<a class="card feature" data-link="paper" href="https://arxiv.org/abs/2602.05843" target="_blank" rel="noopener">
<div class="ico">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.6" aria-hidden="true">
<path d="M14 3H6a2 2 0 0 0-2 2v14a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V9Z" stroke-linejoin="round" />
<path d="M14 3v6h6M8 13h8M8 17h6" stroke-linecap="round" stroke-linejoin="round" />
</svg>
</div>
<h3>Paper</h3>
<p>The full description of the four environments, the two suites and the analysis, on arXiv.</p>
</a>
<a class="card feature" data-link="code" href="https://github.com/xufangzhi/Odyssey-Arena" target="_blank" rel="noopener">
<div class="ico">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.6" aria-hidden="true">
<path d="m8 6-6 6 6 6M16 6l6 6-6 6M14 3l-4 18" stroke-linecap="round" stroke-linejoin="round" />
</svg>
</div>
<h3>Code</h3>
<p>The environments, the task generators and the evaluation harness.</p>
</a>
<a class="card feature" data-link="benchmark" href="https://github.com/xufangzhi/Odyssey-Arena/tree/main/test_data" target="_blank" rel="noopener">
<div class="ico">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.6" aria-hidden="true">
<ellipse cx="12" cy="6" rx="8" ry="3" />
<path d="M4 6v6c0 1.66 3.58 3 8 3s8-1.34 8-3V6M4 12v6c0 1.66 3.58 3 8 3s8-1.34 8-3v-6" />
</svg>
</div>
<h3>Benchmark</h3>
<p>The OdysseyArena-Lite and OdysseyArena-Challenge task instances, in the code repository.</p>
</a>
<a class="card feature" href="#results">
<div class="ico">
<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.6" aria-hidden="true">
<path d="M8 21h8M12 17v4M5 3h14v5a7 7 0 0 1-14 0V3Z" stroke-linejoin="round" />
<path d="M5 5H3v2a3 3 0 0 0 3 3M19 5h2v2a3 3 0 0 1-3 3" stroke-linejoin="round" />
</svg>
</div>
<h3>Leaderboard</h3>
<p>Every evaluated model on the four environments, sortable.</p>
</a>
</div>
</div>
</section>
<!-- ================================================================== -->
<!-- BibTeX -->
<!-- ================================================================== -->
<section class="band-tight" id="cite">
<div class="wrap">
<div class="section-head single">
<span class="eyebrow">Citation</span>
<h2 class="section-title">BibTeX</h2>
</div>
<div class="bibtex rise">
<div class="bibtex-bar">
<span class="dots" aria-hidden="true"><i></i><i></i><i></i></span>
<span class="fn">odysseyarena.bib</span>
<button class="copy-btn" id="copyBib" type="button">
<svg aria-hidden="true" width="13" height="13" viewBox="0 0 16 16" fill="none" stroke="currentColor" stroke-width="1.5">
<rect x="5.5" y="5.5" width="9" height="9" rx="1.6" />
<path d="M10.5 5.5v-3a1 1 0 0 0-1-1h-7a1 1 0 0 0-1 1v7a1 1 0 0 0 1 1h3" />
</svg>
<span>Copy</span>
</button>
</div>
<pre id="bibtex">@article{xu2026odysseyarena,
title={OdysseyArena: Benchmarking Large Language Models For Long-Horizon, Active and Inductive Interactions},
author={Xu, Fangzhi and Yan, Hang and Sun, Qiushi and Wu, Jinyang and Huang, Zixian and Huang, Muye and Gong, Jingyang and Ding, Zichen and Cheng, Kanzhi and Wang, Yian and Che, Xinyu and Sun, Zeyi and Zhang, Jian and Yin, Zhangyue and Luo, Haoran and Kao, Ben and Lin, Qika},
journal={arXiv preprint arXiv:2602.05843},
year={2026}
}</pre>
</div>
</div>
</section>
</main>
<!-- ==================================================================== -->
<!-- Footer -->
<!-- ==================================================================== -->
<footer class="site">
<div class="wrap">
<div class="cols">
<div>
<div class="brand-line">
<img src="public/logo.svg" alt="">
<span>OdysseyArena</span>
</div>
<p style="font-size:14.5px;color:var(--text-dim);max-width:380px">
Environments whose rules are hidden, and a benchmark for the agents that have to find them.
</p>
</div>
<div>
<h2>Project</h2>
<ul>
<li><a href="#overview">Overview</a></li>
<li><a href="#environments">Environments</a></li>
<li><a href="#benchmark">Benchmark</a></li>
<li><a href="#results">Results</a></li>
<li><a href="#analysis">Analysis</a></li>
<li><a href="#cite">Cite</a></li>
</ul>
</div>
<div>
<h2>Artifacts</h2>
<ul>
<li><a href="https://arxiv.org/abs/2602.05843" target="_blank" rel="noopener">Paper</a></li>
<li><a href="https://github.com/xufangzhi/Odyssey-Arena" target="_blank" rel="noopener">Code</a></li>
<li><a href="https://github.com/xufangzhi/Odyssey-Arena/tree/main/test_data" target="_blank" rel="noopener">Benchmark</a></li>
<li><a href="#results">Leaderboard</a></li>
</ul>
</div>
</div>
<div class="fine">
<span>© 2026 The OdysseyArena authors.</span>
<span>Page content licensed
<a href="https://creativecommons.org/licenses/by/4.0/" target="_blank" rel="noopener">CC BY 4.0</a>.</span>
</div>
</div>
</footer>
<script src="static/js/odyssey-tables.js"></script>
<script src="static/js/odyssey-curves.js"></script>
<script src="static/js/odyssey-data.js"></script>
<script src="static/js/lights.js"></script>
<script src="static/js/hero.js"></script>
<script src="static/js/envs.js"></script>
<script src="static/js/charts.js"></script>
<script src="static/js/leaderboard.js"></script>
<script src="static/js/main.js"></script>
<script src="static/js/page.js"></script>
</body>
</html>