-
Notifications
You must be signed in to change notification settings - Fork 65
Expand file tree
/
Copy pathindex.html
More file actions
713 lines (659 loc) · 51.6 KB
/
index.html
File metadata and controls
713 lines (659 loc) · 51.6 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
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
<!DOCTYPE html>
<html lang="zh-CN">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Youtu-Tip 端侧AI助手</title>
<script src="https://cdn.tailwindcss.com"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/gsap/3.12.2/gsap.min.js"></script>
<script src="https://cdnjs.cloudflare.com/ajax/libs/gsap/3.12.2/ScrollTrigger.min.js"></script>
<style>
body { background: radial-gradient(circle at 20% 20%, #f3e8ff 0, #e0f2fe 35%, #ffffff 80%); color: #0f172a; font-family: -apple-system, sans-serif; overflow-x: hidden; }
::-webkit-scrollbar { width: 0px; background: transparent; }
.gradient-text { background: linear-gradient(135deg, #8b5cf6 0%, #3b82f6 60%, #38bdf8 100%); -webkit-background-clip: text; -webkit-text-fill-color: transparent; }
.glass-nav { background: rgba(255, 255, 255, 0.85); backdrop-filter: blur(18px); -webkit-backdrop-filter: blur(18px); border-bottom: 1px solid rgba(15,23,42,0.06); box-shadow: 0 4px 30px rgba(15, 23, 42, 0.08); }
.btn-gradient { background: linear-gradient(135deg, #8b5cf6, #3b82f6); color: #fff; box-shadow: 0 10px 25px rgba(79, 70, 229, 0.25); }
/* 图片叠加样式 */
.img-stack {
position: absolute;
top: 0; left: 0;
width: 100%; height: 100%;
object-fit: cover;
opacity: 0; /* 默认隐藏 */
transition: transform 0.5s ease;
}
/* 第一张默认显示 */
#img-1 { opacity: 1; }
/* 表格样式 */
.data-table th { text-align: left; padding: 12px; color: #475569; font-weight: 600; border-bottom: 1px solid #e2e8f0; font-size: 0.9rem; background: #f8fafc; }
.data-table td { padding: 12px; border-bottom: 1px solid #e2e8f0; font-variant-numeric: tabular-nums; font-size: 0.95rem; color: #0f172a; }
.data-table tr:hover { background: #f1f5f9; }
.highlight-win { color: #2563eb; font-weight: 700; }
.highlight-second { text-decoration: underline; text-decoration-color: #2563eb; }
/* 隐藏类 */
.hidden-panel { display: none; }
</style>
</head>
<body class="antialiased">
<!-- 导航栏 -->
<nav class="fixed top-0 w-full z-50 glass-nav py-4">
<div class="max-w-7xl mx-auto px-6 flex justify-between items-center">
<div class="flex items-center gap-3">
<img src="youtu-llm/assets/logo.svg" alt="Tip Logo" class="h-9 w-9 rounded-lg shadow-sm object-contain">
<span class="font-bold text-xl tracking-tight text-slate-900">Youtu-Tip</span>
</div>
<div class="flex gap-6 text-sm font-medium text-slate-600 items-center">
<a href="#features" class="hover:text-slate-900 transition" data-i18n="nav_features">特性</a>
<a href="#llm" class="hover:text-slate-900 transition" data-i18n="nav_llm">模型</a>
<a href="#framework" class="hover:text-slate-900 transition" data-i18n="nav_framework">框架</a>
<a href="https://github.com/TencentCloudADP/youtu-tip" target="_blank" class="hover:text-slate-900 transition">GitHub</a>
<button id="lang-btn" onclick="toggleLanguage()" class="border border-slate-300 rounded-full px-3 py-1 text-xs hover:bg-slate-100 text-slate-700 transition">English</button>
</div>
</div>
</nav>
<!-- Hero 区域 -->
<header class="relative min-h-screen flex flex-col justify-center items-center text-center px-4 pt-20">
<div class="space-y-6 z-10 max-w-4xl">
<div class="flex flex-col md:flex-row items-center justify-center gap-6 opacity-0 hero-anim mb-8">
<!-- 【需替换】大 Logo -->
<h1 class="text-5xl md:text-7xl font-bold tracking-tight gradient-text" data-i18n="hero_title">
<img src="youtu-tip/docs/assets/model.png" class="inline-block h-[1em] md:h-[1.2em] align-middle drop-shadow-xl">
Youtu-Tip 端侧AI助手
</h1>
</div>
<p class="text-xl md:text-2xl text-slate-600 max-w-2xl mx-auto opacity-0 hero-anim" data-i18n="hero_subtitle">
Tip 是一个主动式端侧 AI 助手，一键调用，智能理解当前工作内容。离线可用，保护隐私。
</p>
<div class="pt-6 opacity-0 hero-anim flex flex-col items-center gap-3">
<a href="https://github.com/TencentCloudADP/youtu-tip/releases" class="btn-gradient px-8 py-3 rounded-full font-semibold text-lg hover:opacity-90 transition transform hover:scale-105 flex items-center gap-2">
<svg class="w-5 h-5" fill="currentColor" viewBox="0 0 20 20"><path d="M13.586 3.586a2 2 0 112.828 2.828l-.793.793-2.828-2.828.793-.793zM11.379 5.793L3 14.172V17h2.828l8.38-8.379-2.83-2.828z"></path></svg>
<span data-i18n="btn_download">下载 macOS 版（M 系列）</span>
</a>
<p class="text-sm text-slate-500" data-i18n="hero_note">目前仅支持 Apple Silicon (M 系列芯片)</p>
</div>
</div>
<!-- 视频嵌入卡片 -->
<div class="mt-16 w-full max-w-6xl mx-auto rounded-2xl overflow-hidden border border-slate-200 bg-white shadow-2xl opacity-0 hero-video-anim">
<div class="relative w-full aspect-video">
<iframe class="absolute inset-0 w-full h-full" src="https://www.youtube.com/embed/c4vczLEmVt4?si=nyCF3zs1UWbzvUrA" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe>
</div>
</div>
</header>
<!-- 滚动视差区域 (修复了高度问题和多图切换) -->
<!-- 注意：这里的 h-screen 是关键，不再写 300vh -->
<section id="features" class="relative h-screen bg-gradient-to-br from-slate-50 via-white to-slate-100 overflow-hidden flex flex-col justify-center border-t border-slate-100">
<div class="max-w-7xl mx-auto w-full px-6 flex flex-col md:flex-row items-center h-full">
<!-- 左侧：文字 -->
<div class="w-full md:w-1/2 relative z-10 h-[300px] flex items-center">
<!-- 特性 1 -->
<div class="feature-text absolute w-full opacity-100 transform translate-y-0" id="txt-1">
<h3 class="text-4xl font-bold text-slate-900 mb-4" data-i18n="feat1_title">一个按键<br><span class="text-slate-500">AI 超级入口</span></h3>
<p class="text-xl text-slate-600" data-i18n="feat1_desc">按下热键或选中文本，Tip 随时待命。打造更智能的 Spotlight，带来极致便捷体验。</p>
</div>
<!-- 特性 2 -->
<div class="feature-text absolute w-full opacity-0 transform translate-y-10" id="txt-2">
<h3 class="text-4xl font-bold text-slate-900 mb-4" data-i18n="feat2_title">端侧模型<br><span class="text-slate-500">百分百隐私</span></h3>
<p class="text-xl text-slate-600" data-i18n="feat2_desc">内置 Youtu-llm 模型，完全离线运行。所有数据仅在本地处理，安全无忧。</p>
</div>
<!-- 特性 3 -->
<div class="feature-text absolute w-full opacity-0 transform translate-y-10" id="txt-3">
<h3 class="text-4xl font-bold text-slate-900 mb-4" data-i18n="feat3_title">读文件、看网页<br><span class="text-slate-500">全能助手</span></h3>
<p class="text-xl text-slate-600" data-i18n="feat3_desc">模拟键鼠操作、接入 MCP 服务器。在本地运行一个真正的多功能智能体。</p>
</div>
<!-- 特性 4 -->
<div class="feature-text absolute w-full opacity-0 transform translate-y-10" id="txt-4">
<h3 class="text-4xl font-bold text-slate-900 mb-4" data-i18n="feat4_title">桌面新技能<br><span class="text-slate-500">一学就掌握</span></h3>
<p class="text-xl text-slate-600" data-i18n="feat4_desc">「技能」机制允许 Tip 学习你教它的方法。定制化属于你的桌面自动化技能。</p>
</div>
</div>
<!-- 右侧：图片堆叠区 -->
<div class="w-full md:w-1/2 flex items-center justify-center pl-0 md:pl-10 mt-10 md:mt-0">
<div class="relative w-full max-w-[540px] aspect-square bg-white rounded-3xl border border-slate-200 overflow-hidden shadow-xl">
<img id="img-1" src="https://github.com/user-attachments/assets/8650e4cf-e6ba-487f-bf62-b1abc4dbbbb7" class="img-stack" alt="Interaction">
<img id="img-2" src="https://github.com/user-attachments/assets/1cd3aab8-6160-4bf5-bce5-a9160b7e0190" class="img-stack" alt="Privacy">
<img id="img-3" src="https://github.com/user-attachments/assets/57b1d07b-34fa-449b-877f-0c1a59eb7180" class="img-stack" alt="Agent">
<img id="img-4" src="https://github.com/user-attachments/assets/a343f80c-b681-4d52-bad7-f85feab9021f" class="img-stack" alt="Skill">
</div>
</div>
</div>
</section>
<!-- 章节：模型介绍 -->
<section id="llm" class="py-32 bg-white border-t border-slate-200">
<div class="max-w-7xl mx-auto px-6">
<!-- 模型类型切换 -->
<div class="flex justify-center mb-12">
<div class="inline-flex bg-slate-100 rounded-full p-1.5 shadow-inner">
<button onclick="switchModelType('llm')" id="model-tab-llm" class="px-6 py-2.5 rounded-full text-sm font-semibold text-white bg-gradient-to-r from-purple-500 to-blue-500 shadow transition-all duration-300">Youtu-LLM</button>
<button onclick="switchModelType('vl')" id="model-tab-vl" class="px-6 py-2.5 rounded-full text-sm font-semibold text-slate-600 hover:text-slate-900 transition-all duration-300">Youtu-VL</button>
</div>
</div>
<!-- Youtu-LLM 面板 -->
<div id="model-panel-llm">
<div class="text-center mb-16 space-y-4">
<span class="text-blue-500 font-semibold tracking-wide text-sm uppercase">The Engine Inside</span>
<h2 class="text-4xl md:text-6xl font-bold text-slate-900">
Youtu-LLM <br>
<span class="text-slate-500 text-3xl md:text-5xl" data-i18n="llm_subtitle">小巧身躯，强悍性能</span>
</h2>
<p class="text-slate-600 max-w-3xl mx-auto text-lg pt-4" data-i18n="llm_intro">
1.96B 参数，128K 上下文。专为 STEM、代码与智能体任务设计。<br>
在轻量级模型中重新定义 SOTA。
</p>
</div>
<!-- 核心亮点 Grid -->
<div class="grid grid-cols-1 md:grid-cols-3 gap-6 mb-24">
<div class="p-8 rounded-3xl bg-white border border-slate-200 shadow-sm hover:border-blue-200 hover:shadow-md transition duration-300">
<div class="text-4xl mb-4">🔬</div>
<h3 class="text-xl font-bold text-slate-900 mb-2" data-i18n="highlight1_title">STEM 优先设计</h3>
<p class="text-slate-600 text-sm leading-relaxed" data-i18n="highlight1_desc">从词表构建到课程学习，一切以提升 STEM 能力和智能体能力为出发点。</p>
</div>
<div class="p-8 rounded-3xl bg-white border border-slate-200 shadow-sm hover:border-blue-200 hover:shadow-md transition duration-300">
<div class="text-4xl mb-4">🤖</div>
<h3 class="text-xl font-bold text-slate-900 mb-2" data-i18n="highlight2_title">原生智能体能力</h3>
<p class="text-slate-600 text-sm leading-relaxed" data-i18n="highlight2_desc">128K 长上下文训练 + Agentic Mid-training，轻松应对多轮复杂端侧任务。</p>
</div>
<div class="p-8 rounded-3xl bg-white border border-slate-200 shadow-sm hover:border-blue-200 hover:shadow-md transition duration-300">
<div class="text-4xl mb-4">🚀</div>
<h3 class="text-xl font-bold text-slate-900 mb-2" data-i18n="highlight3_title">SOTA 性能 (MLA)</h3>
<p class="text-slate-600 text-sm leading-relaxed" data-i18n="highlight3_desc">基于 dense MLA 架构，超越传统 GQA/MHA 范式，轻松集成现有生态。</p>
</div>
</div>
<!-- 性能对比面板 -->
<div class="bg-white rounded-3xl border border-slate-200 overflow-hidden shadow-lg">
<div class="p-6 border-b border-slate-200 flex flex-col md:flex-row justify-between items-center gap-4">
<h3 class="text-xl font-bold text-slate-900" data-i18n="bench_title">性能基准测试</h3>
<div class="flex bg-slate-100 rounded-lg p-1">
<button onclick="switchTab('base')" id="tab-base" class="px-4 py-1.5 rounded-md text-sm font-medium text-white bg-gradient-to-r from-purple-500 to-blue-500 shadow transition">Base Model</button>
<button onclick="switchTab('instruct')" id="tab-instruct" class="px-4 py-1.5 rounded-md text-sm font-medium text-slate-600 hover:text-slate-900 transition">Instruct Model</button>
</div>
</div>
<div class="p-6 overflow-x-auto">
<!-- Base Model Table -->
<div id="panel-base" class="animate-fade">
<table class="w-full data-table text-slate-700">
<thead>
<tr>
<th>Benchmark</th>
<th class="text-blue-600">Youtu-LLM-2B</th>
<th>Qwen3-1.7B</th>
<th>SmoLM3-3B</th>
<th>Gemma3-4B</th>
<!-- <th>Qwen3-4B</th> -->
<th>Llama3.1-8B</th>
</tr>
</thead>
<tbody>
<tr><td>MMLU-Pro (Commonsense)</td><td class="highlight-win">48.4%</td><td>34.9%</td><td>35.3%</td><td>29.4%</td><td>36.2%</td></tr>
<tr><td>GSM8K (STEM)</td><td class="highlight-win">77.6%</td><td>68.2%</td><td>67.3%</td><td>38.5%</td><td>47.8%</td></tr>
<tr><td>MBPP+ (Coding)</td><td class="highlight-win">81.8%</td><td>71.0%</td><td>66.1%</td><td>61.9%</td><td>62.7%</td></tr>
<tr><td>HumanEval (Coding)</td><td class="highlight-win">64.6%</td><td>49.9%</td><td>34.8%</td><td>36.6%</td><td>36.0%</td></tr>
<tr><td>LiveCodeBench v6</td><td class="highlight-win">9.7%</td><td>5.1%</td><td>2.9%</td><td>2.9%</td><td>3.4%</td></tr>
<tr><td>SWE-Bench-Verified (Agent)</td><td class="highlight-win">17.7%</td><td>0.6%</td><td>7.2%</td><td>5.7%</td><td>-</td></tr>
</tbody>
</table>
<p class="text-xs text-slate-500 mt-4 text-center">* 部分数据摘录，查看 <a href="https://github.com/TencentCloudADP/youtu-tip/tree/master/youtu-llm" class="text-blue-600 hover:underline">README</a> 获取完整复现代码。</p>
</div>
<!-- Instruct Model Table -->
<div id="panel-instruct" class="hidden-panel animate-fade">
<table class="w-full data-table text-slate-700">
<thead>
<tr>
<th>Benchmark</th>
<th class="text-blue-600">Youtu-LLM-2B</th>
<th>Qwen3-1.7B</th>
<th>SmoLM3-3B</th>
<!-- <th>Qwen3-4B</th> -->
</tr>
</thead>
<tbody>
<tr><td>GAIA (Agent)</td><td class="highlight-win">33.9%</td><td>11.4%</td><td>11.7%</td></tr>
<tr><td>xbench (Agent)</td><td class="highlight-win">19.5%</td><td>11.7%</td><td>13.9%</td></tr>
<tr><td>SWE-Bench-Verified</td><td class="highlight-win">17.7%</td><td>0.6%</td><td>7.2%</td></tr>
<tr><td>EnConda-Bench</td><td class="highlight-win">21.5%</td><td>10.8%</td><td>3.5%</td></tr>
<tr><td>τ²-Bench (Tool)</td><td class="highlight-win">15.0%</td><td>2.6%</td><td>9.7%</td></tr>
<!-- <tr><td>GAIA (Agent)</td><td class="highlight-win">33.9%</td><td>11.4%</td><td>11.7%</td><td>25.5%</td></tr>
<tr><td>xbench (Agent)</td><td class="highlight-win">19.5%</td><td>11.7%</td><td>13.9%</td><td>18.4%</td></tr>
<tr><td>SWE-Bench-Verified</td><td class="highlight-win">17.7%</td><td>0.6%</td><td>7.2%</td><td>5.7%</td></tr>
<tr><td>EnConda-Bench</td><td class="highlight-win">21.5%</td><td>10.8%</td><td>3.5%</td><td>16.1%</td></tr>
<tr><td>τ²-Bench (Tool)</td><td class="highlight-win">15.0%</td><td>2.6%</td><td>9.7%</td><td>10.9%</td></tr> -->
</tbody>
</table>
<p class="text-xs text-slate-500 mt-4 text-center">* 对比同级别指令模型表现。</p>
</div>
</div>
</div>
</div>
<!-- Youtu-LLM 面板结束 -->
<!-- Youtu-VL 面板 -->
<div id="model-panel-vl" class="hidden-panel">
<div class="text-center mb-16 space-y-4">
<span class="text-purple-500 font-semibold tracking-wide text-sm uppercase">Vision-Language Model</span>
<h2 class="text-4xl md:text-6xl font-bold text-slate-900">
Youtu-VL <br>
<span class="text-slate-500 text-3xl md:text-5xl" data-i18n="vl_subtitle">视觉语言，统一智能</span>
</h2>
<p class="text-slate-600 max-w-3xl mx-auto text-lg pt-4" data-i18n="vl_intro">
基于 Youtu-LLM 构建的 4B 参数视觉语言模型。首创 VLUAS 范式，<br>
标准 VLM 架构实现视觉任务，无需额外模块。
</p>
</div>
<!-- VL 核心亮点 Grid -->
<div class="grid grid-cols-1 md:grid-cols-2 gap-6 mb-24">
<div class="p-8 rounded-3xl bg-white border border-slate-200 shadow-sm hover:border-purple-200 hover:shadow-md transition duration-300">
<div class="text-4xl mb-4">🔗</div>
<h3 class="text-xl font-bold text-slate-900 mb-2" data-i18n="vl_highlight1_title">VLUAS 统一监督</h3>
<p class="text-slate-600 text-sm leading-relaxed" data-i18n="vl_highlight1_desc">视觉-语言统一自回归监督，将视觉信号转化为自回归目标。联合重建视觉与文本 token，保留密集视觉信息的同时增强多模态理解。</p>
</div>
<div class="p-8 rounded-3xl bg-white border border-slate-200 shadow-sm hover:border-purple-200 hover:shadow-md transition duration-300">
<div class="text-4xl mb-4">👁️</div>
<h3 class="text-xl font-bold text-slate-900 mb-2" data-i18n="vl_highlight2_title">无需额外模块</h3>
<p class="text-slate-600 text-sm leading-relaxed" data-i18n="vl_highlight2_desc">标准 VLM 架构原生支持视觉任务：分割、深度估计、定位、检测。单一模型，灵活应对各种视觉-语言需求。</p>
</div>
</div>
<!-- VL 性能对比面板 -->
<div class="bg-white rounded-3xl border border-slate-200 overflow-hidden shadow-lg">
<div class="p-6 border-b border-slate-200 flex flex-col md:flex-row justify-between items-center gap-4">
<h3 class="text-xl font-bold text-slate-900" data-i18n="vl_bench_title">性能基准测试</h3>
<div class="flex bg-slate-100 rounded-lg p-1">
<button onclick="switchVLTab('vision')" id="vl-tab-vision" class="px-4 py-1.5 rounded-md text-sm font-medium text-white bg-gradient-to-r from-purple-500 to-blue-500 shadow transition" data-i18n="vl_tab_vision">视觉任务</button>
<button onclick="switchVLTab('general')" id="vl-tab-general" class="px-4 py-1.5 rounded-md text-sm font-medium text-slate-600 hover:text-slate-900 transition" data-i18n="vl_tab_general">多模态任务</button>
</div>
</div>
<div class="p-6 overflow-x-auto">
<!-- Vision-Centric Table -->
<div id="vl-panel-vision" class="animate-fade">
<table class="w-full data-table text-slate-700 text-sm">
<thead>
<tr>
<th>Benchmark</th>
<!-- <th>Qwen3-VL 4B</th> -->
<th>InternVL-3.5 4B</th>
<th>UFO 8B</th>
<th class="text-purple-600">Youtu-VL 4B</th>
</tr>
</thead>
<tbody>
<tr><td colspan="5" class="font-semibold bg-slate-50 text-slate-700">Visual Grounding</td></tr>
<tr><td>RefCOCO val</td><td>92.5%</td><td>91.8%</td><td class="highlight-win">93.6%</td></tr>
<tr><td>RefCOCO testA</td><td>94.3%</td><td>94.3%</td><td class="highlight-win">95.2%</td></tr>
<tr><td>RefCOCO+ val</td><td>87.6%</td><td>86.9%</td><td class="highlight-win">90.1%</td></tr>
<tr><td>RefCOCOg test</td><td>89.3%</td><td>88.6%</td><td class="highlight-win">92.9%</td></tr>
<tr><td colspan="5" class="font-semibold bg-slate-50 text-slate-700">Semantic Segmentation</td></tr>
<tr><td>ADE20k</td><td>×</td><td>54.5%</td><td class="highlight-second">54.2%</td></tr>
<tr><td>Cityscapes</td><td>×</td><td>-</td><td class="highlight-win">70.4%</td></tr>
<tr><td>VOC20</td><td>×</td><td>-</td><td class="highlight-win">92.5%</td></tr>
<tr><td colspan="5" class="font-semibold bg-slate-50 text-slate-700">Referring Segmentation</td></tr>
<tr><td>RefCOCO val</td><td>×</td><td>80.0%</td><td class="highlight-win">80.7%</td></tr>
<tr><td>RefCOCO testA</td><td>×</td><td>81.6%</td><td class="highlight-win">82.0%</td></tr>
<tr><td colspan="5" class="font-semibold bg-slate-50 text-slate-700">Depth Estimation</td></tr>
<tr><td>NYUv2 (δ1)</td><td>×</td><td>93.6%</td><td class="highlight-second">90.4%</td></tr>
<tr><td>Cityscapes</td><td>×</td><td>-</td><td class="highlight-win">92.7%</td></tr>
<tr><td colspan="5" class="font-semibold bg-slate-50 text-slate-700">Object Counting</td></tr>
<tr><td>TallyQA-Simple</td><td>77.6%</td><td>×</td><td class="highlight-win">85.1%</td></tr>
<tr><td>CountBench</td><td>79.4%</td><td>×</td><td class="highlight-win">88.6%</td></tr>
</tbody>
</table>
<p class="text-xs text-slate-500 mt-4 text-center" data-i18n="vl_table_note1">* × 表示模型不支持该任务。</p>
</div>
<!-- General Multimodal Table -->
<div id="vl-panel-general" class="hidden-panel animate-fade">
<table class="w-full data-table text-slate-700 text-sm">
<thead>
<tr>
<th>Benchmark</th>
<th>Qwen3-VL 8B</th>
<th>InternVL-3.5 4B</th>
<!-- <th>Qwen3-VL 4B</th> -->
<th class="text-purple-600">Youtu-VL 4B</th>
</tr>
</thead>
<tbody>
<tr><td colspan="5" class="font-semibold bg-slate-50 text-slate-700">General VQA</td></tr>
<tr><td>MMBench_EN</td><td>84.5%</td><td>80.3%</td><td class="highlight-second">83.9%</td></tr>
<tr><td>MMStar</td><td>70.9%</td><td>65.0%</td><td class="highlight-win">71.1%</td></tr>
<tr><td>MME (/2800)</td><td>-</td><td>2272</td><td class="highlight-win">2384</td></tr>
<tr><td>ScienceQA_val</td><td>-</td><td>-</td><td class="highlight-win">97.0%</td></tr>
<tr><td colspan="5" class="font-semibold bg-slate-50 text-slate-700">Multimodal Reasoning</td></tr>
<tr><td>VisuLogic</td><td>22.5%</td><td>-</td>><td class="highlight-win">25.7%</td></tr>
<tr><td>MathVista_mini</td><td>77.2%</td><td>77.1%</td><td class="highlight-second">76.5%</td></tr>
<tr><td>MathVerse_mini</td><td>62.1%</td><td>45.8%</td><td class="highlight-second">56.5%</td></tr>
<tr><td>VLMsAreBlind</td><td>74.0%</td><td>-</td><td class="highlight-win">88.9%</td></tr>
<tr><td colspan="5" class="font-semibold bg-slate-50 text-slate-700">OCR & Document</td></tr>
<tr><td>AI2D_test</td><td>85.7%</td><td>82.6%</td><td class="highlight-second">85.6%</td></tr>
<tr><td>DocVQA_val</td><td>96.1%</td><td>92.4%</td><td class="highlight-second">94.4%</td></tr>
<tr><td>ChartQA_test</td><td>89.6%</td><td>86.0%</td><td class="highlight-second">85.3%</td></tr>
<tr><td colspan="5" class="font-semibold bg-slate-50 text-slate-700">GUI Agent</td></tr>
<tr><td>ScreenSpot Pro</td><td>54.6%</td><td>-</td><td class="highlight-win">59.6%</td></tr>
<tr><td>OSWorld</td><td>33.9%</td><td>-</td><td class="highlight-win">38.8%</td></tr>
<tr><td colspan="5" class="font-semibold bg-slate-50 text-slate-700">Real-World</td></tr>
<tr><td>RealWorldQA</td><td>71.5%</td><td>66.3%</td><td class="highlight-win">74.6%</td></tr>
</tbody>
</table>
<p class="text-xs text-slate-500 mt-4 text-center" data-i18n="vl_table_note2">* 对比同级别 VLM 指令模型表现。</p>
</div>
</div>
</div>
</div>
<!-- Youtu-VL 面板结束 -->
</div>
</section>
<!-- 章节：Youtu-Agent 框架 -->
<section id="framework" class="py-32 bg-[#F5F5F7] border-t border-slate-200 relative overflow-hidden">
<div class="absolute -top-[20%] -right-[10%] w-[800px] h-[800px] bg-blue-100/50 blur-[100px] rounded-full pointer-events-none mix-blend-multiply"></div>
<div class="absolute top-[20%] -left-[10%] w-[600px] h-[600px] bg-purple-100/50 blur-[100px] rounded-full pointer-events-none mix-blend-multiply"></div>
<div class="max-w-7xl mx-auto px-6 relative z-10">
<div class="flex flex-col md:flex-row gap-16 items-center">
<!-- 左侧：文案与特性 -->
<div class="w-full md:w-1/2 space-y-10">
<div class="space-y-4">
<span class="inline-block px-3 py-1 rounded-full bg-blue-100 text-blue-600 font-semibold tracking-wide text-xs uppercase mb-2">Open Source Framework</span>
<h2 class="text-4xl md:text-5xl font-bold text-slate-900 leading-tight">
Build with <br>
<span class="text-transparent bg-clip-text bg-gradient-to-r from-blue-600 to-violet-600">Youtu-Agent</span>
</h2>
<p class="text-slate-600 text-lg leading-relaxed" data-i18n="agent_intro">
Youtu-Tip 的强大动力源泉。一个灵活、高性能的智能体构建框架。支持自动化生成、混合策略优化，助你轻松构建 SOTA 级应用。
</p>
</div>
<!-- 特性列表 -->
<div class="space-y-6">
<!-- Feature 1 -->
<div class="flex gap-5 items-start">
<div class="flex-shrink-0 w-12 h-12 rounded-2xl bg-white shadow-sm border border-slate-200 flex items-center justify-center text-2xl text-blue-500">
⚡️
</div>
<div>
<h4 class="text-slate-900 font-bold text-lg" data-i18n="agent_feat1_title">自动化智能体生成</h4>
<p class="text-slate-500 text-sm mt-1 leading-relaxed" data-i18n="agent_feat1_desc">描述需求，Meta-Agent 自动编写工具代码与配置。</p>
</div>
</div>
<!-- Feature 2 -->
<div class="flex gap-5 items-start">
<div class="flex-shrink-0 w-12 h-12 rounded-2xl bg-white shadow-sm border border-slate-200 flex items-center justify-center text-2xl text-purple-500">
🧠
</div>
<div>
<h4 class="text-slate-900 font-bold text-lg" data-i18n="agent_feat2_title">无需训练的持续进化</h4>
<p class="text-slate-500 text-sm mt-1 leading-relaxed" data-i18n="agent_feat2_desc">独创 Training-Free GRPO，低成本积累经验，越用越强。</p>
</div>
</div>
<!-- Feature 3 -->
<div class="flex gap-5 items-start">
<div class="flex-shrink-0 w-12 h-12 rounded-2xl bg-white shadow-sm border border-slate-200 flex items-center justify-center text-2xl text-amber-500">
🏆
</div>
<div>
<h4 class="text-slate-900 font-bold text-lg" data-i18n="agent_feat3_title">SOTA 性能验证</h4>
<p class="text-slate-500 text-sm mt-1 leading-relaxed" data-i18n="agent_feat3_desc">WebWalkerQA (71.47%) 与 GAIA (72.8%) 领跑者。</p>
</div>
</div>
</div>
<div class="pt-6">
<a href="https://github.com/TencentCloudADP/youtu-agent" target="_blank" class="inline-flex items-center gap-2 text-slate-900 bg-white border border-slate-300 px-6 py-3 rounded-full hover:bg-slate-50 hover:shadow-md transition-all group font-medium">
<svg class="w-5 h-5 text-slate-900" fill="currentColor" viewBox="0 0 24 24"><path d="M12 0c-6.626 0-12 5.373-12 12 0 5.302 3.438 9.8 8.207 11.387.599.111.793-.261.793-.577v-2.234c-3.338.726-4.033-1.416-4.033-1.416-.546-1.387-1.333-1.756-1.333-1.756-1.089-.745.083-.729.083-.729 1.205.084 1.839 1.237 1.839 1.237 1.07 1.834 2.807 1.304 3.492.997.107-.775.418-1.305.762-1.604-2.665-.305-5.467-1.334-5.467-5.931 0-1.311.469-2.381 1.236-3.221-.124-.303-.535-1.524.117-3.176 0 0 1.008-.322 3.301 1.23.957-.266 1.983-.399 3.003-.404 1.02.005 2.047.138 3.006.404 2.291-1.552 3.297-1.23 3.297-1.23.653 1.653.242 2.874.118 3.176.77.84 1.235 1.911 1.235 3.221 0 4.609-2.807 5.624-5.479 5.921.43.372.823 1.102.823 2.222v3.293c0 .319.192.694.801.576 4.765-1.589 8.199-6.086 8.199-11.386 0-6.627-5.373-12-12-12z"/></svg>
<span data-i18n="btn_view_repo">View on GitHub</span>
<span class="group-hover:translate-x-1 transition-transform text-slate-400">→</span>
</a>
</div>
</div>
<!-- 右侧：代码终端演示 -->
<div class="w-full md:w-1/2">
<div class="rounded-2xl bg-[#0F1117] border border-slate-900/10 shadow-2xl overflow-hidden font-mono text-sm leading-relaxed relative group transform transition hover:-translate-y-1 duration-500">
<!-- Mac 窗口控制点 -->
<div class="bg-[#1e1e1e] px-4 py-3 border-b border-white/5 flex gap-2 items-center">
<div class="flex gap-2">
<div class="w-3 h-3 rounded-full bg-[#ff5f56]"></div>
<div class="w-3 h-3 rounded-full bg-[#ffbd2e]"></div>
<div class="w-3 h-3 rounded-full bg-[#27c93f]"></div>
</div>
<div class="ml-auto text-xs text-slate-400 font-medium opacity-60">research_agent.yaml</div>
</div>
<!-- 代码内容 -->
<div class="p-6 text-gray-300 overflow-x-auto">
<div class="space-y-1 font-mono text-sm">
<div><span class="text-purple-400">agent</span>:</div>
<div class="pl-4"><span class="text-blue-400">name</span>: research_agent</div>
<div class="pl-4"><span class="text-blue-400">instructions</span>: <span class="text-green-400">"You are a helpful research assistant..."</span></div>
<div class="mt-2"><span class="text-purple-400">env</span>:</div>
<div class="pl-4"><span class="text-blue-400">name</span>: e2b</div>
<div class="pl-4"><span class="text-blue-400">config</span>: {}</div>
<div class="mt-2"><span class="text-purple-400">context_manager</span>:</div>
<div class="pl-4"><span class="text-blue-400">name</span>: base</div>
<div class="pl-4"><span class="text-blue-400">config</span>: {}</div>
<div class="mt-2"><span class="text-purple-400">toolkits</span>:</div>
<div class="pl-4"><span class="text-blue-400">search</span>:</div>
<div class="pl-8"><span class="text-blue-400">activated_tools</span>: [<span class="text-green-400">"search"</span>, <span class="text-green-400">"web_qa"</span>]</div>
<div class="pl-4"><span class="text-blue-400">python_executor</span>:</div>
<div class="pl-8"><span class="text-blue-400">activated_tools</span>: [<span class="text-green-400">"execute_python_code"</span>]</div>
</div>
</div>
<!-- 悬浮提示 -->
<div class="absolute bottom-4 right-4 bg-blue-600/20 text-blue-300 px-3 py-1 rounded text-xs border border-blue-500/30 opacity-0 group-hover:opacity-100 transition">
Sample Configuration
</div>
</div>
</div>
</div>
</div>
</section>
<!-- 下载区域 -->
<footer class="py-12 text-center text-slate-500 text-sm bg-white border-t border-slate-200">
<p>© 2025 Youtu-Tip. Open Source Project.</p>
</footer>
<!-- 脚本逻辑 -->
<script>
gsap.registerPlugin(ScrollTrigger);
// --- 1. Hero 动画 ---
const tl = gsap.timeline();
tl.to(".hero-anim", { y: 0, opacity: 1, duration: 1, stagger: 0.2, ease: "power3.out" })
.to(".hero-video-anim", { opacity: 1, y: 0, duration: 1, ease: "power2.out" }, "-=0.5");
// --- 2. 核心修复：滚动视差 (文字+图片切换) ---
// 我们不设置 CSS 高度，而是让 ScrollTrigger 钉住(pin) 3000px 的滚动距离
// 这样滚动完 3000px 后，页面会自动解冻，不会留下空白
const scrollTl = gsap.timeline({
scrollTrigger: {
trigger: "#features",
start: "top top",
end: "+=4000", // 滚动距离：决定动画播放时长
pin: true, // 钉住
scrub: true, // 绑定滚动条
anticipatePin: 1
}
});
scrollTl
// 阶段 1 -> 2
.to("#txt-1", { opacity: 0, y: -20, duration: 1 }) // 文字1 上浮消失
.to("#img-1", { opacity: 0, duration: 1 }, "<") // 图片1 淡出
.to("#txt-2", { opacity: 1, y: 0, duration: 1 }) // 文字2 上浮出现
.to("#img-2", { opacity: 1, duration: 1 }, "<") // 图片2 淡入
.to({}, { duration: 1 }) // 停顿一会儿，让用户看完内容
// 阶段 2 -> 3
.to("#txt-2", { opacity: 0, y: -20, duration: 1 })
.to("#img-2", { opacity: 0, duration: 1 }, "<")
.to("#txt-3", { opacity: 1, y: 0, duration: 1 })
.to("#img-3", { opacity: 1, duration: 1 }, "<")
.to({}, { duration: 1 }) // 停顿
// 阶段 3 -> 4
.to("#txt-3", { opacity: 0, y: -20, duration: 1 })
.to("#img-3", { opacity: 0, duration: 1 }, "<")
.to("#txt-4", { opacity: 1, y: 0, duration: 1 })
.to("#img-4", { opacity: 1, duration: 1 }, "<")
.to({}, { duration: 1 }); // 停顿
// 3. 模型类型切换逻辑 (LLM vs VL)
function switchModelType(type) {
const llmBtn = document.getElementById('model-tab-llm');
const vlBtn = document.getElementById('model-tab-vl');
const llmPanel = document.getElementById('model-panel-llm');
const vlPanel = document.getElementById('model-panel-vl');
if (type === 'llm') {
llmBtn.className = "px-6 py-2.5 rounded-full text-sm font-semibold text-white bg-gradient-to-r from-purple-500 to-blue-500 shadow transition-all duration-300";
vlBtn.className = "px-6 py-2.5 rounded-full text-sm font-semibold text-slate-600 hover:text-slate-900 transition-all duration-300";
llmPanel.style.display = 'block';
vlPanel.style.display = 'none';
} else {
vlBtn.className = "px-6 py-2.5 rounded-full text-sm font-semibold text-white bg-gradient-to-r from-purple-500 to-blue-500 shadow transition-all duration-300";
llmBtn.className = "px-6 py-2.5 rounded-full text-sm font-semibold text-slate-600 hover:text-slate-900 transition-all duration-300";
vlPanel.style.display = 'block';
llmPanel.style.display = 'none';
}
}
// 4. VL Tab 切换逻辑
function switchVLTab(type) {
const visionBtn = document.getElementById('vl-tab-vision');
const generalBtn = document.getElementById('vl-tab-general');
const visionPanel = document.getElementById('vl-panel-vision');
const generalPanel = document.getElementById('vl-panel-general');
if (type === 'vision') {
visionBtn.className = "px-4 py-1.5 rounded-md text-sm font-medium text-white bg-gradient-to-r from-purple-500 to-blue-500 shadow transition";
generalBtn.className = "px-4 py-1.5 rounded-md text-sm font-medium text-slate-600 hover:text-slate-900 transition";
visionPanel.style.display = 'block';
generalPanel.style.display = 'none';
} else {
generalBtn.className = "px-4 py-1.5 rounded-md text-sm font-medium text-white bg-gradient-to-r from-purple-500 to-blue-500 shadow transition";
visionBtn.className = "px-4 py-1.5 rounded-md text-sm font-medium text-slate-600 hover:text-slate-900 transition";
generalPanel.style.display = 'block';
visionPanel.style.display = 'none';
}
}
// 5. LLM Tab 切换逻辑
function switchTab(type) {
const baseBtn = document.getElementById('tab-base');
const instructBtn = document.getElementById('tab-instruct');
const basePanel = document.getElementById('panel-base');
const instructPanel = document.getElementById('panel-instruct');
if (type === 'base') {
baseBtn.className = "px-4 py-1.5 rounded-md text-sm font-medium text-white bg-gradient-to-r from-purple-500 to-blue-500 shadow transition";
instructBtn.className = "px-4 py-1.5 rounded-md text-sm font-medium text-slate-600 hover:text-slate-900 transition";
basePanel.style.display = 'block';
instructPanel.style.display = 'none';
} else {
instructBtn.className = "px-4 py-1.5 rounded-md text-sm font-medium text-white bg-gradient-to-r from-purple-500 to-blue-500 shadow transition";
baseBtn.className = "px-4 py-1.5 rounded-md text-sm font-medium text-slate-600 hover:text-slate-900 transition";
instructPanel.style.display = 'block';
basePanel.style.display = 'none';
}
}
// 4. 语言切换 (包含新章节)
const translations = {
zh: {
nav_features: "特性",
nav_llm: "模型",
nav_framework: "框架",
nav_download: "下载",
hero_title: "Youtu-Tip 端侧AI助手",
hero_subtitle: "Tip 是一个主动式端侧 AI 助手，一键调用，智能理解当前工作内容。离线可用，保护隐私。",
btn_download: "下载 macOS 版（M 系列）",
hero_note: "目前仅支持 Apple Silicon (M 系列芯片)",
feat1_title: '一个按键<br><span class="text-slate-500">AI 超级入口</span>',
feat1_desc: "按下热键或选中文本，Tip 随时待命。打造更智能的 Spotlight，带来极致便捷体验。",
feat2_title: '端侧模型<br><span class="text-slate-500">百分百隐私</span>',
feat2_desc: "内置 Youtu-llm 模型，完全离线运行。所有数据仅在本地处理，安全无忧。",
feat3_title: '读文件、看网页<br><span class="text-slate-500">全能助手</span>',
feat3_desc: "模拟键鼠操作、接入 MCP 服务器。在本地运行一个真正的多功能智能体。",
feat4_title: '桌面新技能<br><span class="text-slate-500">一学就掌握</span>',
feat4_desc: "「技能」机制允许 Tip 学习你教它的方法。定制化属于你的桌面自动化技能。",
// LLM Section
llm_subtitle: "小巧身躯，强悍性能",
llm_intro: "1.96B 参数，128K 上下文。专为 STEM、代码与智能体任务设计。<br>在轻量级模型中重新定义 SOTA。",
highlight1_title: "STEM 优先设计",
highlight1_desc: "从词表构建到课程学习，一切以提升 STEM 能力和智能体能力为出发点。",
highlight2_title: "原生智能体能力",
highlight2_desc: "128K 长上下文训练 + Agentic Mid-training，轻松应对多轮复杂端侧任务。",
highlight3_title: "SOTA 性能 (MLA)",
highlight3_desc: "基于 dense MLA 架构，超越传统 GQA/MHA 范式，轻松集成现有生态。",
bench_title: "性能基准测试",
// Youtu-VL Section
vl_subtitle: "视觉语言，统一智能",
vl_intro: "基于 Youtu-LLM 构建的 4B 参数视觉语言模型。首创 VLUAS 范式，<br>标准 VLM 架构实现视觉任务，无需额外模块。",
vl_highlight1_title: "VLUAS 统一监督",
vl_highlight1_desc: "视觉-语言统一自回归监督，将视觉信号转化为自回归目标。联合重建视觉与文本 token，保留密集视觉信息的同时增强多模态理解。",
vl_highlight2_title: "无需额外模块",
vl_highlight2_desc: "标准 VLM 架构原生支持视觉任务：分割、深度估计、定位、检测。单一模型，灵活应对各种视觉-语言需求。",
vl_bench_title: "性能基准测试",
vl_tab_vision: "视觉任务",
vl_tab_general: "多模态任务",
vl_table_note1: "* × 表示模型不支持该任务。",
vl_table_note2: "* 对比同级别 VLM 指令模型表现。",
// Youtu-Agent Section
agent_intro: "Youtu-Tip 的强大动力源泉。一个灵活、高性能的智能体构建框架。支持自动化生成、混合策略优化，助你轻松构建 SOTA 级应用。",
agent_feat1_title: "自动化智能体生成",
agent_feat1_desc: "描述需求，Meta-Agent 自动编写工具代码与智能体配置。",
agent_feat2_title: "无需训练的持续进化",
agent_feat2_desc: "独创 Training-Free GRPO，低成本积累经验，越用越强。",
agent_feat3_title: "SOTA 性能验证",
agent_feat3_desc: "WebWalkerQA (71.47%) 与 GAIA (72.8%) 领跑者。",
btn_view_repo: "在 GitHub 上查看",
// End
ready_title: "准备好体验未来了吗？",
btn_github: "前往 GitHub Release 下载"
},
en: {
nav_features: "Features",
nav_llm: "Model",
nav_framework: "Framework",
nav_download: "Download",
hero_title: "Youtu-Tip AI Assistant",
hero_subtitle: "Tip is a proactive on-device AI assistant: one hotkey, understands what's on your screen. Offline-ready and privacy-first.",
btn_download: "Download for macOS (Apple Silicon)",
hero_note: "Currently Apple Silicon (M-Series) only",
feat1_title: 'One Keystroke<br><span class="text-slate-500">Super AI Entry</span>',
feat1_desc: "Press a hotkey or select text, Tip is ready. A smarter Spotlight for the ultimate convenient experience.",
feat2_title: 'On-Device Model<br><span class="text-slate-500">100% Privacy</span>',
feat2_desc: "Built-in Youtu-llm models running completely offline. All data is processed locally.",
feat3_title: 'Reads Files & Web<br><span class="text-slate-500">All-Round Agent</span>',
feat3_desc: "Simulates mouse/keyboard actions, connects to MCP servers. Run a versatile agent locally.",
feat4_title: 'New Desktop Skills<br><span class="text-slate-500">Learn and Master</span>',
feat4_desc: "\"Skill\" mechanism allows Tip to learn from methods you teach it. Customize your own desktop automation skills.",
// LLM Section
llm_subtitle: "Small Size, Massive Power",
llm_intro: "1.96B Params, 128K Context. Designed for STEM, Coding & Agent tasks.<br>Redefining SOTA for lightweight models.",
highlight1_title: "STEM-First Design",
highlight1_desc: "From vocabulary construction to curriculum learning, everything starts with improving STEM and Agent capabilities.",
highlight2_title: "Native Agent Capabilities",
highlight2_desc: "128K long context training + Agentic Mid-training. Easily handles multi-turn complex on-device tasks.",
highlight3_title: "SOTA Performance (MLA)",
highlight3_desc: "Based on dense MLA architecture, surpassing traditional GQA/MHA paradigms. Easy to integrate.",
bench_title: "Benchmarks",
// Youtu-VL Section
vl_subtitle: "Unified Vision-Language Intelligence",
vl_intro: "A 4B-parameter Vision-Language Model built on Youtu-LLM. Pioneering VLUAS paradigm,<br>enabling vision-centric tasks with a standard VLM architecture—no task-specific modules needed.",
vl_highlight1_title: "VLUAS Unified Supervision",
vl_highlight1_desc: "Vision-Language Unified Autoregressive Supervision transforms visual signals into autoregressive targets. Jointly reconstructing visual and text tokens preserves dense visual information while enhancing multimodal understanding.",
vl_highlight2_title: "No Extra Modules Needed",
vl_highlight2_desc: "Standard VLM architecture natively supports vision tasks: segmentation, depth estimation, grounding, detection. One model, flexible for all vision-language requirements.",
vl_bench_title: "Benchmarks",
vl_tab_vision: "Vision Tasks",
vl_tab_general: "Multimodal Tasks",
vl_table_note1: "* × indicates the model does not support this task.",
vl_table_note2: "* Compared with same-tier VLM instruct models.",
// Youtu-Agent Section
agent_intro: "The engine behind Youtu-Tip. A flexible, high-performance framework for autonomous agents. Featuring automated generation and hybrid policy optimization.",
agent_feat1_title: "Automated Agent Generation",
agent_feat1_desc: "Describe tasks, and Meta-Agent auto-generates tool code & agent configs.",
agent_feat2_title: "Continuous Evolution",
agent_feat2_desc: "Training-Free GRPO enables experience learning without parameter updates.",
agent_feat3_title: "Verified SOTA Performance",
agent_feat3_desc: "Leading scores on WebWalkerQA (71.47%) and GAIA (72.8%).",
btn_view_repo: "View on GitHub",
// End
ready_title: "Ready to experience the future?",
btn_github: "Download from GitHub Release"
}
};
let currentLang = 'zh';
function toggleLanguage() {
currentLang = currentLang === 'zh' ? 'en' : 'zh';
const btn = document.getElementById('lang-btn');
// 更新按钮文字
btn.textContent = currentLang === 'zh' ? 'English' : '中文';
// 更新所有带有 data-i18n 的元素
document.querySelectorAll('[data-i18n]').forEach(el => {
const key = el.getAttribute('data-i18n');
if (translations[currentLang][key]) {
el.innerHTML = translations[currentLang][key];
}
});
}
</script>
</body>
</html>