update gh-pages

Author: hexuan21

index.html

@@ -163,8 +163,8 @@ <h2 class="subtitle is-3 publication-subtitle"> Think before You Score in Genera
                   <sup>6</sup><span class="author-block">Yi Lu,</span>
                   <sup>2</sup><span class="author-block">Keming Wu,</span>
                   <sup>2</sup><span class="author-block">Benjamin Schneider,</span>
-                  <sup>2</sup><span class="author-block">Quy Duc Do,</span>
                   <br>
+                  <sup>2</sup><span class="author-block">Quy Duc Do,</span>
                   <sup>2</sup><span class="author-block">Zhuofeng Li,</span>
                   <sup>6</sup><span class="author-block">Yiming Jia,</span>
                   <sup>2</sup><span class="author-block">Yuxuan Zhang,</span>
@@ -189,6 +189,7 @@ <h2 class="subtitle is-3 publication-subtitle"> Think before You Score in Genera
                     <sup>3</sup>Independent, 
                     <sup>4</sup>2077AI, 
                     <sup>5</sup>M-A-P, 
+                    <br>
                     <sup>6</sup>University of Toronto, 
                     <sup>7</sup>Zhejiang University, 
                     <sup>8</sup>Abaka AI,
@@ -198,7 +199,7 @@ <h2 class="subtitle is-3 publication-subtitle"> Think before You Score in Genera
             <br>   
             <div class="is-size-5 publication-authors">
               <span class="author-block">
-                    *Equal Contribution, Xuan He lead the project
+                    *Equal Contribution, Xuan leads the project
               </span>
             </div>
             <div class="is-size-5 publication-authors">
@@ -365,56 +366,67 @@ <h2 class="title is-4">
 
   <div class="hero-body has-text-centered">
     <h2 class="title is-4">
-      <span style="vertical-align: middle">Evaluation Results</span>
+      <span style="vertical-align: middle">Evaluation on VideoScore2-Bench</span>
     </h2>
   </div>
   <div class="container is-max-desktop">
     <div class="columns is-centered">
       <div class="column is-full-width">
-        <h2 class="title is-5"><span style="font-size: 100%;">
-          VideoScore2-Bench</span></h2>
         <p>
           We test VideoScore2 on the test set of VideoFeedback2, containing 500 videos with human scores from three dimensions. 
           Below we show the results of some scoring/reward models for image or video, like VideoReward, VisionReward, Q-Insight, DeQA-Score, and MLLM-prompting methods like Claude-Sonnet-4, GPT-5, Gemini-2.5-Pro, etc and our VideoScore2. 
         </p>
-        <img id="res_vs2_bench" width="100%" src="static/images/res_vs2_bench.png">
+        <img id="res_vs2_bench" width="70%" src="static/images/res_vs2_bench.png">
+      </div>
+    </div>
+  </div>
+
+
 
-        <br>
-        <h2 class="title is-5"><span style="font-size: 100%;">
-          <img id="painting_icon" width="3%" src="static/images/ec_icon.png">
-          Out-of-Domain Benchmark</span></h2>
+  <div class="hero-body has-text-centered">
+    <h2 class="title is-4">
+      <span style="vertical-align: middle">Evaluation on Out-of-Domain Benchmarks</span>
+    </h2>
+  </div>
+  <div class="container is-max-desktop">
+    <div class="columns is-centered">
+      <div class="column is-full-width">
         <p>
-          We further test on four out-of-domain (OOD) benchmarks: two pairwise preference and two point
-score. Preference benchmark results include ties. As shown in Table below, while VideoScore2 is
+          We further test on four out-of-domain (OOD) benchmarks: two pairwise preference (VideoGenReward-Bench and T2VQA-DB (preference version)) and two point score (MJ-Bench-Video and Video-Phy2-test). 
+
+Preference benchmark results include ties. As shown in Table below, while VideoScore2 is
  not always the top model on each benchmark, it achieves the highest overall average.
         </p>
-        <img id="res_ood_bench" width="100%" src="static/images/res_ood_bench.png">
+        <img id="res_ood_bench" width="70%" src="static/images/res_ood_bench.png">
       </div>
     </div>
   </div>
 
+  <div class="hero-body has-text-centered">
+    <h2 class="title is-4">
+      <span style="vertical-align: middle">Best-of-N Sampling with VideoScore2</span>
+    </h2>
+  </div>
+
   <div class="container is-max-desktop">
     <div class="columns is-centered">
       <div class="column is-full-width">
-        <h2 class="title is-5"><span style="font-size: 100%;">
-          Best-of-N Sampling</span></h2>
         <p>
           We evaluate VIDEOSCORE2 with best-of-n (BoN) sampling (n = 5), where the model selects the
  best video among candidates. Six T2V models of moderate or poor quality are used, avoiding very
  strong ones to highlight the BoN effect. For 500 prompts, each model generates 500 × 5 videos.
  Comparison on VBench shows BoN consistently outperforms random sampling, confirm
 ing that VIDEOSCORE2 effectively guides higher-quality selection.
         </p>
-        <img id="BoN" width="100%" src="static/images/BoN.png">
+        <img id="BoN" width="70%" src="static/images/BoN.png">
       </div>
     </div>
   </div>
+    
+</section>
 
 
-  
 
-    
-</section>
 
 <section class="hero is-light is-small">
   <div class="hero-body has-text-centered">
@@ -424,9 +436,9 @@ <h2 class="title is-3">
   </div>
 </section>  
 
-<img id="cs1" width="100%" src="static/images/case1.png">
-<img id="cs2" width="100%" src="static/images/case2.png">
-<img id="cs3" width="100%" src="static/images/case3.png">
+<img id="cs1" width="70%" src="static/images/case1.png">
+<img id="cs2" width="70%" src="static/images/case2.png">
+<img id="cs3" width="70%" src="static/images/case3.png">
 
 
 <section class="section" id="BibTeX">