Vector Space Visualization 3D

1. Token Embedding: Each word converts to a high-dimensional vector.

2. Context Aggregation: Model combines token vectors with weighted importance.

3. Vector Arithmetic: Prediction = Σ(w_i × v_i) with exponential weights.

4. Similarity Search: Finds closest vocabulary token via cosine similarity.

5. Connection Opacity: Lines fade based on similarity AND distance.