Upload README.md with huggingface_hub

README.md

@@ -1,14 +1,12 @@
 ---
 title: World Model Demo
 emoji: 🧠
-colorFrom: green
-colorTo: purple
+colorFrom: blue
+colorTo: green
 sdk: gradio
-sdk_version: 4.44.0
+sdk_version: 4.44.1
 app_file: app.py
 pinned: false
-license: mit
-short_description: Interactive demo of how AI agents learn to dream and plan
 tags:
   - world-model
   - reinforcement-learning
@@ -17,80 +15,20 @@ tags:
   - cognitive-architecture
 ---
 
-# 🧠 World Model Demo
+# World Model Demo
 
-An interactive visualization demonstrating how intelligent agents build internal models of their world to plan actions without trial-and-error in reality.
+Interactive demonstration of world model concepts in AI planning and decision-making.
 
-## The Concept
+## Features
+- **Visual Grid Environment**: 8x8 grid with obstacles and goals
+- **Phase-Based Learning**: Observe → Plan → Act → Learn cycle
+- **Real-time Statistics**: Track predictions, errors, and model confidence
+- **Educational Overlays**: See how the agent predicts and plans
 
-A **World Model** is an internal simulation that an agent uses to predict the outcomes of its actions. This is how both biological and artificial intelligence can "think ahead" - imagining consequences before committing to actions.
+## Concepts Demonstrated
+- Model-based reinforcement learning (MuZero, Dreamer)
+- World state representation and prediction
+- Planning with learned dynamics models
+- The imagination-execution loop
 
-### The Three Phases
-
-| Phase | What Happens | Real-World Analogy |
-|-------|--------------|-------------------|
-| 🔍 **Exploration** | Random movement to discover physics rules | A baby learning to crawl by bumping into things |
-| 💭 **Dreaming** | Planning using *only* the internal model | Mentally rehearsing a speech before giving it |
-| 🚀 **Execution** | Following the imagined plan | Actually performing the rehearsed speech |
-
-## How to Use
-
-1. **Configure** the grid size and obstacle pattern
-2. **Explore** - Watch the agent learn the world's physics through random movement
-3. **Dream** - See the agent plan a path using only its learned model (no real movement!)
-4. **Execute** - Watch the plan work in reality
-
-Or just click **Run All Phases** to see the complete demonstration!
-
-## Technical Details
-
-The World Model is implemented as a simple dictionary:
-```python
-transitions[(state, action)] = next_state
-```
-
-During the **Dreaming** phase, the agent uses BFS search through this dictionary - it never calls the real environment! This is the key insight: planning happens entirely in the agent's "imagination."
-
-## Why This Matters
-
-This demo illustrates the foundation of modern AI systems:
-
-- **MuZero** (DeepMind) - Learned world models for game playing
-- **Dreamer** - World models for robot control
-- **PlaNet** - Planning with learned dynamics
-- **Human cognition** - We constantly simulate futures in our minds
-
-## Architecture
-
-```
-┌─────────────────────────────────────────────────────────────┐
-│                    INTELLIGENT AGENT                        │
-│  ┌─────────────┐    ┌─────────────┐    ┌─────────────┐     │
-│  │ ENVIRONMENT │◄──►│ WORLD MODEL │◄──►│  PLANNING   │     │
-│  │  (Reality)  │    │  (Memory)   │    │  (Dreams)   │     │
-│  └─────────────┘    └─────────────┘    └─────────────┘     │
-│         │                  ▲                   │            │
-│         │    learn()       │      predict()    │            │
-│         └──────────────────┴───────────────────┘            │
-└─────────────────────────────────────────────────────────────┘
-```
-
-## Legend
-
-- 🤖 Agent position
-- ⭐ Goal location
-- 🏁 Start position
-- 🧱 Wall/obstacle
-- 🟢 Green border = States the model has learned
-- 🟣 Purple dashed = States searched during planning
-- 🔵 Cyan = Planned path
-
-## References
-
-- Ha, D., & Schmidhuber, J. (2018). World Models
-- Hafner, D., et al. (2019). Dream to Control: Learning Behaviors by Latent Imagination
-- Schrittwieser, J., et al. (2020). Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model
-
----
-
-*Built with ❤️ using Gradio*
+Built for the Anthropic Research Fellowship application.