Model
A trained function from input to output. Umbrella term — an LLM, an embedding model, an OCR engine, and a classifier are all models.
Last updated
In ML, a “model” is a trained function from input to output. “The model” in everyday speech could mean an LLM, an embedding model, a classifier, an OCR engine, or a tiny decision tree. The only thing they share is that their weights were learned from data — which is what distinguishes a model from a rule-based system.
Why it matters for your research. When someone says “a model says X”, it’s worth asking which model, trained on what, doing which task. Conflating an embedding model (maps text → vector) with a generative model (maps text → more text) produces the most common misconceptions in DH discussions of AI.
In Archēglyph. Every artefact the UI shows records the id of the model that produced it, so the answer to “which model?” is always one hover away. See provenance.
Not to be confused with. LLM is one specific kind of model, not a synonym for “model”.