YuyaoTechnology
/

MatterSeek

@@ -45,21 +45,21 @@ license_link: LICENSE
   <tr><td>Laboratory Synthesis</td><td>Our in-house laboratory runs automated experiments 24/7; real experimental data is collected and fed back for online model training, continuously optimizing model performance.</td></tr>
   <tr><td>Characterization & Validation</td><td> XRD phase analysis and SEM/EDS morphology and composition analysis, comparing measured values against predicted values.</td></tr>
 </table>
-<h2>🎬 视频演示</h2>
-<h3>宣传视频（60s）</h3>
 <video controls width="100%" poster="assets/promo_cover.jpg">
   <source src="https://huggingface.co/YuyaoTechnology/MatterSeek/resolve/main/%E5%85%AC%E5%8F%B8-%E4%BF%AE%E6%94%B92.mp4" type="video/mp4">
 </video>
-<h3>产品操作演示（3min）</h3>
 <video controls width="100%" poster="assets/demo_cover.jpg">
   <source src="https://huggingface.co/YuyaoTechnology/MatterSeek/resolve/main/%E4%BA%A7%E5%93%81%E4%BD%BF%E7%94%A8-2.mp4" type="video/mp4">
 </video>
-<blockquote><b>演示路径</b>：登录 → 自然语言指令/目标性能 → 元素约束筛选 → 多维可视化 → 导出配方/工艺 → 提交线下验证</blockquote>
 <h2>Model Architecture</h2>
 <table>
   <tr><th>Item</th><th>Description</th></tr>
   <tr><td>Architecture</td><td>Proprietary HNN hierarchical neural network (physics-reasoning-driven). Two-tier design: a shared backbone learns cross-domain material representations from the 1,035-d fingerprint; on top sit 192 second-level expert sub-models grouped under 17 first-level vertical domains, each fine-tuned on materials that exhibit the target property. Physics-informed regularization embeds symmetry, conservation laws and thermodynamic constraints as soft penalties, so the network outputs physically plausible predictions rather than pure interpolation.</td></tr>
-  <tr><td>材料描述符</td><td>1035 种</td></tr>
   <tr><td>Domain sub-models</td><td>192 second-level (二级) performance-variable sub-models organized under 17 first-level (一级) vertical domains: superconductivity, magnetocaloric, magneto-transport, magnetism, catalysis & surfaces, electronic transport, optics, nuclear & irradiation, mechanics, crystal structure, thermoelectrics, thermodynamics, thermal transport, phonons, ferroelectrics, barocaloric, and fabrication & cost. Each sub-model is a dedicated predictor fine-tuned on materials that exhibit the target property. See category breakdown below.</td></tr>
   <tr><td>Input formats</td><td>Natural-language instructions, property targets, composition formulas.</td></tr>
   <tr><td>Output formats</td><td>Predicted properties with uncertainty bounds, recommended compositions (Pareto-front candidates), processing parameters (sintering / annealing windows), and ranked experiment proposals.</td></tr>

   <tr><td>Laboratory Synthesis</td><td>Our in-house laboratory runs automated experiments 24/7; real experimental data is collected and fed back for online model training, continuously optimizing model performance.</td></tr>
   <tr><td>Characterization & Validation</td><td> XRD phase analysis and SEM/EDS morphology and composition analysis, comparing measured values against predicted values.</td></tr>
 </table>
+<h2>Video Demos</h2>
+<h3>Promo Video</h3>
 <video controls width="100%" poster="assets/promo_cover.jpg">
   <source src="https://huggingface.co/YuyaoTechnology/MatterSeek/resolve/main/%E5%85%AC%E5%8F%B8-%E4%BF%AE%E6%94%B92.mp4" type="video/mp4">
 </video>
+<h3>Product Walkthrough</h3>
 <video controls width="100%" poster="assets/demo_cover.jpg">
   <source src="https://huggingface.co/YuyaoTechnology/MatterSeek/resolve/main/%E4%BA%A7%E5%93%81%E4%BD%BF%E7%94%A8-2.mp4" type="video/mp4">
 </video>
+<blockquote><b>Suggested demo path:</b>Sign in → enter a natural-language instruction or target properties → filter by elemental constraints → explore multi-dimensional result visualization → export composition / processing parameters → submit for offline validation</blockquote>
 <h2>Model Architecture</h2>
 <table>
   <tr><th>Item</th><th>Description</th></tr>
   <tr><td>Architecture</td><td>Proprietary HNN hierarchical neural network (physics-reasoning-driven). Two-tier design: a shared backbone learns cross-domain material representations from the 1,035-d fingerprint; on top sit 192 second-level expert sub-models grouped under 17 first-level vertical domains, each fine-tuned on materials that exhibit the target property. Physics-informed regularization embeds symmetry, conservation laws and thermodynamic constraints as soft penalties, so the network outputs physically plausible predictions rather than pure interpolation.</td></tr>
+  <tr><td>Material descriptors</td><td>1035</td></tr>
   <tr><td>Domain sub-models</td><td>192 second-level (二级) performance-variable sub-models organized under 17 first-level (一级) vertical domains: superconductivity, magnetocaloric, magneto-transport, magnetism, catalysis & surfaces, electronic transport, optics, nuclear & irradiation, mechanics, crystal structure, thermoelectrics, thermodynamics, thermal transport, phonons, ferroelectrics, barocaloric, and fabrication & cost. Each sub-model is a dedicated predictor fine-tuned on materials that exhibit the target property. See category breakdown below.</td></tr>
   <tr><td>Input formats</td><td>Natural-language instructions, property targets, composition formulas.</td></tr>
   <tr><td>Output formats</td><td>Predicted properties with uncertainty bounds, recommended compositions (Pareto-front candidates), processing parameters (sintering / annealing windows), and ranked experiment proposals.</td></tr>