See how random 4D vectors become meaningful through training (100 sentences)
What You're Watching
The task: Predict which words appear near each other in sentences.
What happens: Words that appear in similar contexts get pulled together.
Animals cluster because they all appear near words like "runs", "eats", "fur".
Fruits cluster because they appear near "sweet", "eat", "grows".
The embeddings start random — watch them organize as training progresses.
Controls
50ms
0
Training Steps
-
Loss (lower = better)
0%
Cluster Quality
2D Projection of 4D Embedding Space
Animals
Fruits
Vehicles
Countries
Raw Vectors (4 Dimensions)
What you're seeing: Each word has 4 numbers (its position in 4D space).
Similar words end up with similar numbers. The bars show each dimension's value from -2 (left) to +2 (right).