Lab: Compare Token Representations

Mission

Run the local Embedding Similarity Lab. Compare the same three tokens under one-hot and supplied dense vectors, reproduce cosine calculations, and export an Embedding Similarity Record.

Guided start

1. Run One-hot vectors with query robot.
2. Record that both different-token cosine values are 0.
3. Switch only the representation to Supplied dense vectors.
4. Run again and preserve every vector, dot product, norm, and cosine.
5. Explain why token spelling stayed fixed while measured similarity changed.

Procedure

1. Preserve both one-hot comparisons.
2. Predict which dense token will be closer to robot.
3. Recalculate both dense cosine values.
4. Compare the sign and magnitude without inventing axis meanings.
5. Preserve your first semantic claim and write a bounded correction.
6. Complete Observation, Measurement, Decision, and Limitation.
7. Export the record and confirm it names ai-for-beginners-embedding-similarity-lab.

Success criteria and boundary

• both vector representations remain visible;
• each dense cosine can be reproduced from its operands;
• representation change is separated from token change;
• the correction follows only supplied geometry; and
• the limitation rejects synonymy, truth, fairness, causation, and other-model claims.

Use only supplied tokens and vectors. The Lab retrieves no text, profiles no person, and contacts no service.