forgot to save the html

index.html

@@ -141,6 +141,14 @@
 
     /* push the hero video down so the button never overlaps */
     .hero > video{ margin-top:.8rem; display:block; max-width:980px; width:100%; border-radius:.8rem; border:var(--border); background:var(--card); }
+    
+    .hint{
+      margin-top:.35rem; font-size:.88rem; color:var(--muted);
+      border-left:3px solid color-mix(in oklab, var(--accent), transparent 65%);
+      padding:.2rem .6rem; line-height:1.4;
+    }
+
+
 
   </style>
 </head>
@@ -223,7 +231,6 @@
                 <th scope="col">Modality</th>
                 <th scope="col">N</th>
                 <th scope="col">Top‑1 Acc (%)</th>
-                <th scope="col">Top‑3 Acc (%)</th>
                 <th scope="col">Updated</th>
               </tr>
             </thead>
@@ -271,11 +278,12 @@
         </figure>
 
         <!-- Group 0: Cross-Modal Gain heatmap -->
-        <div class="results-group"
+        <div class="results-group">
           <h3>Cross-Modal Gain (CMG)</h3>
             <figure class="figure-card">
-              <img src="XModel_gain-2.png" alt="Cross-Modal Gain heatmap" loading="lazy">
-              <figcaption>CMG across tasks and models.</figcaption>
+              <img src="xmod_cis.png" alt="Cross-Modal Gain heatmap" loading="lazy">
+              <figcaption>CMG in percentage points per engine; horizontal bars are 95\% paired-bootstrap CIs on the paired clip set.</figcaption>
+              <div class="hint">Look for positive values: these mean AV was better than either audio or video alone. Gains usually appear for outcome prediction, but rarely for object or material recognition.</div>
             </figure>
         </div>
 
@@ -286,35 +294,32 @@
             <figure class="figure-card">
               <img src="Heatmap_Audio.png" alt="Average audio weight across models" loading="lazy">
               <figcaption>Audio weight by model.</figcaption>
+              <div class="hint">What to look for: Red = model relies more on audio, Blue = model relies less. 
+                Engines that “listen” more may gain on outcome prediction, but not always.</div>
             </figure>
             <figure class="figure-card">
               <img src="Heatmap_Video.png" alt="Average video weight across models" loading="lazy">
               <figcaption>Video weight by model.</figcaption>
+              <div class="hint">What to look for: Red = model relies more on video, Blue = model relies less. 
+                Engines that “look” more often ignore sound, which can explain weak cross-modal gains.</div>
             </figure>
           </div>
         </div>
 
-        <!-- Group 2: Top-1 accuracy per task -->
+        <!-- Group 2: Top-1 Accuracy: single combined figure -->
         <div class="media-card full-span">
           <div class="results-group">
             <h3>Top-1 accuracy by task</h3>
-            <div class="grid-3">
-              <figure class="figure-card">
-                <img src="Accuracy_plot_material.png" alt="Top-1 accuracy for object" loading="lazy">
-                <figcaption>Object.</figcaption>
-              </figure>
-              <figure class="figure-card">
-                <img src="Accuracy_plot_object.png" alt="Top-1 accuracy for material" loading="lazy">
-                <figcaption>Material.</figcaption>
-              </figure>
-              <figure class="figure-card">
-                <img src="Accuracy_plot_outcome.png" alt="Top-1 accuracy for outcome" loading="lazy">
-                <figcaption>Outcome.</figcaption>
-              </figure>
-            </div>
+            <figure class="figure-card">
+              <img src="accuracy_all.png" alt="Top-1 accuracy per model across object, material, and outcome for A, V, AV" loading="lazy">
+              <figcaption>Top-1 accuracy with 95% CIs (A, V, AV) across tasks and models.</figcaption>
+              <div class="hint">What to look for: V is usually highest; AV improves over A and sometimes nudges past V on outcome. 
+                Big gaps A→AV mean sound is helpful; AV≈V means little extra benefit.</div>
+            </figure>
           </div>
         </div>
 
+
     </section>
 
     <section id="contact" class="section">