Skip to contents

Predict binary probabilities from a fitted probe

Usage

# S3 method for class 'llm_probe'
predict(object, newdata, layer = NULL, ...)

Arguments

object

An llm_probe result.

newdata

A relm_trace with the same component and neuron coordinates and one position per prompt. It can contain new prompt IDs and needs no labels.

layer

Optional fitted layer index. NULL uses the layer selected on development CV, regardless of held-out performance.

...

Ignored.

Value

A numeric probability vector for the positive class, named by prompt ID in increasing order. Uses the saved development-only transformation and fit.

Examples

# See llm_probe() for a complete model-free fit and prediction example.