feat: update doc

app.py

@@ -123,11 +123,14 @@ with gr.Blocks(
     with gr.Column(elem_classes="container"):
         gr.Markdown(
             """
-            # Interactive Image Generation
+            # Interactive Visualization of a Latent Program Network (LPN)
             
-            ## Method Overview
-            This interactive demo showcases our novel image generation method that uses coordinate-based control.
-            The process allows precise control over generated patterns through a coordinate-conditioning mechanism.
+            ## Introduction
+            The LPN is an architecture for inductive program synthesis that builds in test-time adaption
+            by learning a latent space that can be used for search.
+            This interactive demo showcases a latent traversal of the LPN in the latent program space.
+            More specifically, the decoder of the LPN is conditioned on a latent vector representing 
+            an abstract program, which is then used to generate an output.
             """
         )
 
@@ -144,12 +147,13 @@ with gr.Blocks(
         gr.Markdown(
             """
             ### How to Use
-            1. Choose a pattern generation task using the radio buttons
-            2. View the target pattern for your selected task
-            3. Click anywhere in the heatmap to specify coordinates in the latent space
-            4. See the generated image based on your selection
+            1. Choose a pattern task using the radio buttons
+            2. View the input-output pairs for your selected task
+            3. The goal is to find the latent that will generate the right third image for the given input
+            4. Click anywhere in the heatmap to specify coordinates in the latent space
+            5. See the generated image based on your selection
             
-            Use the "Find Optimal Latent" button to automatically select pre-determined optimal coordinates.
+            Use the "Find Optimal Latent" button to find the latent that maximizes likelihood of generating the other input-output pairs.
             """
         )
 
@@ -167,9 +171,9 @@ with gr.Blocks(
                         interactive=True,
                     )
 
-                gr.Markdown("### Coordinate Selector")
+                gr.Markdown("### Latent Space Search")
                 gr.Markdown(
-                    "Click anywhere in the image below to select (x, y) coordinates in the latent space"
+                    "Click anywhere in the 2D latent space below to condition the decoder on a specific latent vector."
                 )
 
                 with gr.Column(elem_classes="coordinate-container"):
@@ -188,7 +192,7 @@ with gr.Blocks(
 
             # Right column for images
             with gr.Column(scale=1):
-                gr.Markdown("### Reference Pattern")
+                gr.Markdown("### Input-Output Pairs")
                 with gr.Column(elem_classes="image-preview-container"):
                     reference_image = gr.Image(
                         value="imgs/pattern_0.png",
@@ -211,8 +215,6 @@ with gr.Blocks(
             gr.Markdown(
                 """
                 ### Technical Details
-                Our approach uses a novel coordinate-conditioning mechanism that allows precise control over the generated patterns. 
-                The heatmap visualization shows the distribution of pattern characteristics across the latent space.
                 
                 For more information, please refer to our [paper](https://arxiv.org/pdf/2411.08706) or GitHub [repository](https://github.com/clement-bonnet/lpn).
                 """