A practical lab · about 17 minutes
Interactive learning lab
How AI works · guide 4 of 8
The semantic vector-space explorer
0/5 core signals
Saved only on this device
Vector lab · BriefingScene 1 / 6
Similarity is geometry after training
A query and document become vectors; distance ranks semantic neighbours.
Similarity is geometry after training ready.
Predict
Inspect
Verify
Embedding models place related text near each other in high-dimensional space. A 2D picture is only a projection, but it explains semantic search, clustering and retrieval—and why exact IDs or logical distinctions need other methods.
Near means similar under the model.
Similarity is not entailment.
Hybrid retrieval covers different failure modes.