Definition
Plain language
A small simple classifier trained on a model's internal states to test what information they contain.
As stated in the literature
A linear classifier trained on frozen intermediate activations to detect whether a particular concept is linearly decodable from the representation.
Also called: linear probing, linear classifier, linear probes, probe, probes
Why it matters: Linear probes are a cheap, standard way to ask 'is this concept actually represented here?' without invasive interventions.
For example, training a one-layer classifier on a transformer's middle layer can reveal whether the model already 'knows' the part of speech of each word at that depth.
Heard on the show
“The only way to solve XOR with a linear classifier is if your features actually contain joint, pairwise information about both inputs together.”Episode 002 — An AI Ran a Real Optics Lab for 21 Hours and Found a Transformer-Shaped Pattern in Light