Update README.md

README.md

@@ -10,4 +10,33 @@ pinned: false
 short_description: Compares the effectiveness of slogans generated using RL.
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Gym Campaign Slogan Comparator
+
+This application compares the effectiveness of slogans generated by two models for a gym campaign. Each model generates slogans along with a predicted effectiveness score. The app calculates the success rate of each model by comparing their slogans pairwise.
+
+## Features
+
+- Compare two lists of slogans with their effectiveness scores.
+- Determine which model performs better overall.
+- Visualize the success rate of each model in a simple interface.
+- Shareable Gradio app for easy use.
+
+## Inputs
+
+- **Model X Slogans:** A list of slogans with scores (format: `slogan,score` per line).
+- **Model Y Slogans:** Same format for the second model.
+
+## Outputs
+
+- **Model X Success Rate:** Percentage of pairwise wins for Model X.
+- **Model Y Success Rate:** Percentage of pairwise wins for Model Y.
+
+## Reinforcement Learning Context
+
+Reinforcement Learning (RL) can be applied to optimize slogan generation over time. In this app:
+
+- Each slogan generation by the models is considered an **action**.
+- The effectiveness score acts as a **reward signal**.
+- By comparing slogans and updating the model to maximize wins (success rate), RL techniques like policy gradients can be used to improve the quality of slogans over iterations.
+
+This app provides a simple interface to visualize these comparisons and can serve as a foundation for RL-based optimization of advertising campaigns.