ml-pokedex a collection of tiny RDF experiments

Four small experiments with language models trained on RDF instead of English, over a single Pokémon knowledge graph. Each runs entirely in your browser. Each draws what the model says, not what the page knows.

My first go at models that read and write the Semantic Web, published as a curiosity. The honest verdict is on the about page: a neural net loses to gzip as a database, but the rules it learns are the interesting part.

recall · 80 MB model

The Memoriser live

Pick real Pokémon. The model recalls their stored facts (typing, abilities, egg group) and draws the constellation, sprites and all. It memorised the whole graph.

pick subjects → draw what it remembers
imagination · 11 MB model

The Minter live

Name a Pokémon that never existed. The model can't look it up, so it invents one: a coherent typing, weaknesses, moves, a habitat, all made up on the spot.

a fresh IRI → a whole new creature
reasoning · same 11 MB model

The Reasoner live

Choose a typing. The model derives its weaknesses from a type chart it was never handed, and the page grades it live against the real chart. It aces the typings it saw and stumbles on the ones held back.

a typing → derived weaknesses, graded
the story · captured samples

The Arc

Scrub through training, epoch by epoch, and watch RDF structure appear in layers: first syntax, then vocabulary, then schema, and reasoning last of all. The RDF tinyshakespeare.

garbage bytes → grammar → meaning
One rule across all four: the page holds names, never relations. It knows that pikachu and electric exist. It does not know Pikachu is Electric. Every fact you see came out of a model. Delete the model files and these pages could still list and lay out subjects, knowing nothing about them.