| <!DOCTYPE html>
|
| <html lang="zh-CN">
|
| <head>
|
| <meta charset="UTF-8">
|
| <meta name="viewport" content="width=device-width, initial-scale=1.0">
|
| <title>数据总量概览:预训练与后训练阶段</title>
|
| <style>
|
| :root {
|
| --primary-color: #2563eb;
|
| --secondary-color: #10b981;
|
| --text-main: #1f2937;
|
| --text-muted: #6b7280;
|
| --bg-color: #ffffff;
|
| --card-bg: #f9fafb;
|
| --border-color: #e5e7eb;
|
| }
|
|
|
| body {
|
| font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, "Helvetica Neue", Arial, sans-serif;
|
| background-color: var(--bg-color);
|
| color: var(--text-main);
|
| margin: 0;
|
| padding: 40px 60px;
|
| box-sizing: border-box;
|
| display: flex;
|
| flex-direction: column;
|
| height: 100vh;
|
| justify-content: center;
|
| }
|
|
|
| h1 {
|
| text-align: center;
|
| font-size: 2.5rem;
|
| margin-bottom: 10px;
|
| color: #111827;
|
| font-weight: 800;
|
| letter-spacing: 2px;
|
| }
|
|
|
| .subtitle {
|
| text-align: center;
|
| color: var(--text-muted);
|
| font-size: 1.2rem;
|
| margin-bottom: 40px;
|
| }
|
|
|
| .container {
|
| display: flex;
|
| gap: 40px;
|
| width: 100%;
|
| max-width: 1400px;
|
| margin: 0 auto;
|
| flex: 1;
|
| }
|
|
|
| .column {
|
| flex: 1;
|
| display: flex;
|
| flex-direction: column;
|
| gap: 20px;
|
| }
|
|
|
| .column-header {
|
| font-size: 1.8rem;
|
| font-weight: bold;
|
| padding-bottom: 15px;
|
| border-bottom: 3px solid;
|
| display: flex;
|
| align-items: center;
|
| gap: 10px;
|
| }
|
|
|
| .pre-train .column-header {
|
| color: var(--primary-color);
|
| border-bottom-color: var(--primary-color);
|
| }
|
|
|
| .post-train .column-header {
|
| color: var(--secondary-color);
|
| border-bottom-color: var(--secondary-color);
|
| }
|
|
|
| .card {
|
| background-color: var(--card-bg);
|
| border: 1px solid var(--border-color);
|
| border-radius: 12px;
|
| padding: 24px;
|
| box-shadow: 0 4px 6px -1px rgba(0, 0, 0, 0.05), 0 2px 4px -1px rgba(0, 0, 0, 0.03);
|
| transition: transform 0.2s ease;
|
| }
|
|
|
| .card:hover {
|
| transform: translateY(-2px);
|
| box-shadow: 0 10px 15px -3px rgba(0, 0, 0, 0.05), 0 4px 6px -2px rgba(0, 0, 0, 0.025);
|
| }
|
|
|
| .card-header {
|
| display: flex;
|
| justify-content: space-between;
|
| align-items: center;
|
| margin-bottom: 16px;
|
| }
|
|
|
| .card-title {
|
| font-size: 1.4rem;
|
| font-weight: bold;
|
| display: flex;
|
| align-items: center;
|
| gap: 8px;
|
| }
|
|
|
| .card-total {
|
| font-size: 1.2rem;
|
| font-weight: 900;
|
| background: #e0f2fe;
|
| color: #0369a1;
|
| padding: 4px 12px;
|
| border-radius: 20px;
|
| }
|
|
|
| .post-train .card-total {
|
| background: #d1fae5;
|
| color: #047857;
|
| }
|
|
|
| .data-list {
|
| list-style: none;
|
| padding: 0;
|
| margin: 0;
|
| }
|
|
|
| .data-list li {
|
| position: relative;
|
| padding-left: 20px;
|
| margin-bottom: 12px;
|
| line-height: 1.6;
|
| color: #374151;
|
| }
|
|
|
| .data-list li::before {
|
| content: "•";
|
| position: absolute;
|
| left: 0;
|
| color: #9ca3af;
|
| font-size: 1.2rem;
|
| }
|
|
|
| .highlight {
|
| font-weight: 600;
|
| color: #111827;
|
| }
|
|
|
| .tag {
|
| display: inline-block;
|
| background-color: #f3f4f6;
|
| border: 1px solid #d1d5db;
|
| color: #4b5563;
|
| font-size: 0.85rem;
|
| padding: 2px 8px;
|
| border-radius: 4px;
|
| margin-left: 6px;
|
| vertical-align: middle;
|
| }
|
|
|
| .method-box {
|
| margin-top: 12px;
|
| padding: 12px;
|
| background-color: #f8fafc;
|
| border-left: 4px solid var(--secondary-color);
|
| border-radius: 0 8px 8px 0;
|
| font-size: 0.95rem;
|
| }
|
|
|
| .arrow {
|
| color: #9ca3af;
|
| margin: 0 6px;
|
| }
|
| </style>
|
| </head>
|
| <body>
|
|
|
| <h1>多模态大模型数据大盘</h1>
|
| <div class="subtitle">Data Volume Overview: Pre-training & Post-training</div>
|
|
|
| <div class="container">
|
| <div class="column pre-train">
|
| <div class="column-header">
|
| <span>⚡ 预训练阶段 (Pre-training)</span>
|
| </div>
|
|
|
| <div class="card">
|
| <div class="card-header">
|
| <div class="card-title">🖼️ T2I (Text-to-Image)</div>
|
| <div class="card-total">~ 812M+</div>
|
| </div>
|
| <ul class="data-list">
|
| <li><span class="highlight">华山落盘数据:800M</span>
|
| <div style="margin-top: 4px; font-size: 0.95rem; color: var(--text-muted);">
|
| 含 240M Qwen3vl4b 重打标 Dense 数据 <span class="tag">平均387 tokens</span>,其余为短标注
|
| </div>
|
| </li>
|
| <li><span class="highlight">Banana 蒸馏数据:2.2M</span> <span class="tag">Gemini标注</span></li>
|
| <li><span class="highlight">自收集开源数据:10M</span> <span class="tag">高质量美学图片</span></li>
|
| </ul>
|
| </div>
|
|
|
| <div class="card">
|
| <div class="card-header">
|
| <div class="card-title">🎵 T2A (Text-to-Audio)</div>
|
| <div class="card-total">56,300 Hours+</div>
|
| </div>
|
| <ul class="data-list">
|
| <li><span class="highlight">Audio 数据:6,300 小时</span> (约 2.3M)</li>
|
| <li><span class="highlight">Speech 数据:50,000 小时</span> (约 33M)</li>
|
| <li><span class="highlight">自收集开源数据</span> (补充来源)</li>
|
| </ul>
|
| </div>
|
|
|
| <div class="card">
|
| <div class="card-header">
|
| <div class="card-title">🎬 T2V (Text-to-Video)</div>
|
| <div class="card-total">~ 100M+</div>
|
| </div>
|
| <div style="margin-bottom: 10px;"><span class="tag">>480p</span> <span class="tag">时长 3~10s</span></div>
|
| <ul class="data-list">
|
| <li><span class="highlight">华山数据:100M</span>
|
| <div style="margin-top: 4px; font-size: 0.95rem; color: var(--text-muted);">
|
| 含 22M 数据经 Gemini 重打标注
|
| </div>
|
| </li>
|
| <li><span class="highlight">Seedance 蒸馏:350K</span></li>
|
| </ul>
|
| </div>
|
|
|
| <div class="card">
|
| <div class="card-header">
|
| <div class="card-title">🎞️ T2AV (Text-to-Audio-Video)</div>
|
| <div class="card-total">2M</div>
|
| </div>
|
| </div>
|
| </div>
|
|
|
| <div class="column post-train">
|
| <div class="column-header">
|
| <span>🎯 后训练阶段 (Post-training)</span>
|
| </div>
|
|
|
| <div class="card">
|
| <div class="card-header">
|
| <div class="card-title">✨ I2I (Image-to-Image)</div>
|
| <div class="card-total">5M</div>
|
| </div>
|
| <ul class="data-list">
|
| <li><span class="highlight">Nano Banana 蒸馏数据:5M</span> <span class="tag">1K 分辨率</span></li>
|
| </ul>
|
| </div>
|
|
|
| <div class="card">
|
| <div class="card-header">
|
| <div class="card-title">📽️ I2V (Image-to-Video)</div>
|
| <div class="card-total">5M</div>
|
| </div>
|
| <ul class="data-list">
|
| <li><span class="highlight">精选图生视频数据:5M</span></li>
|
| </ul>
|
| </div>
|
|
|
| <div class="card">
|
| <div class="card-header">
|
| <div class="card-title">🎮 交互数据 (Interactive)</div>
|
| <div class="card-total">1M</div>
|
| </div>
|
| <ul class="data-list">
|
| <li><span class="highlight">游戏交互控制数据:1M</span></li>
|
| </ul>
|
| <div class="method-box">
|
| <strong>🔧 构造链路:</strong><br>
|
| YouTube游戏视频 <span class="arrow">➔</span> VGGt估计姿态 <span class="arrow">➔</span> 提取控制信号
|
| </div>
|
| </div>
|
| </div>
|
| </div>
|
|
|
| </body>
|
| </html> |