SOURCE-LINKED INTELLIGENCE

K/V-Cache Interventions Dissociate Representation Alignment from Persona Expression in Decoder-Only Language Models

arXiv · AI, language, vision and robotics · article · Sep 10, 2026 · UTC

We study K/V-cache interventions -- transplanting a target-conditioned K/V trajectory into a source-persona generation -- as a structured surface for persona control in decoder-only language models. Across 13 intervention configurations applied to Llama-3.1-8B for a fixed source-to-target persona pair, we report two consistent dissociations between representation-level alignment and behavioral expression, plus a common failure under position perturbations. First, all layer-band K/V replacements (early, mid, late) achieve strong local V-space alignment (V-gap 0.91, 0.89, 0.84), but only mid-lay

Read original source ↗ Open in workspace

recordType

paper

region

Global