Follow up: GPT-2's vocabulary as a hyperbolic tree — 32,070 tokens in a Poincaré ball you can fly through [P]

GPT-2's vocabulary as a hyperbolic tree: 32,070 tokens inside a Poincaré ball that you can explore.

Link : https://aethereos.net/static/tinny66666.html

Link named after a reddit user disappointed with my 2D projection ...

It uses the same data as the flat map, GPT-2-small's raw token embeddings and nothing else, but lays them out in hyperbolic space, where tree structures naturally fit.

It runs on your phone. Drag to rotate, pinch to zoom, and tap any token to bring it to the center as the entire space shifts around it. This is a Möbius translation, the natural way to move through hyperbolic geometry. Tap neighbouring tokens to keep exploring.

Why hyperbolic? The vocabulary's similarity structure forms a forest: one giant tree with about 2,300 tokens, a few hundred smaller family trees, and around 6,700 isolated tokens with no close relatives. Trees don't fit well in flat space, but they embed naturally in hyperbolic space, where available room grows exponentially with distance from the center. No optimisation or training is involved. The layout is constructed exactly.