SOURCE-LINKED INTELLIGENCE

How Do Language Models Choose Between Context and Memory?

arXiv · AI, language, vision and robotics · article · Sep 1, 2026 · UTC

When contextual information conflicts with the knowledge stored in model parameters, activation directions can be used to decode and steer which source the model follows. However, steering along a direction does not establish causality: whether the unedited model would naturally use that direction or whether the direction is reusable across tasks. We test these distinctions through counterfactual experiments in unambiguous settings. First, we estimate authority directions from agreement prompts, in which the context and parametric knowledge support the same answer. We then interchange naturall

Read original source ↗ Open in workspace

recordType

paper

region

Global