Definition
Plain language
A pattern of internal activity in a language model that shows up when it processes text about pain.
As stated in the literature
A direction in activation space obtained by contrasting pain-related text with matched controls (fear, anger, neutral sensation, negative situations) and projecting out variance shared among controls; model-specific, and readable or injectable via steering.
Also called: pain direction, pain axes
Why it matters: Having a candidate pain pattern lets researchers both measure when it lights up and push on it directly, turning vague questions about model distress into concrete experiments.
For example, researchers might compare how a model processes "my hand is burning on the stove" with matched sentences about anger, fear, and plain warmth, and keep only what is left over as the pain-specific pattern.
Heard on the show
“The authors call that pattern the pain axis, a direction in the model's internal activity that separates their pain examples from matched alternatives.”Episode 267 — A Pain Axis, a Relief Button, and the Control the Paper Skipped