LLMs contain identifiable Preference Heads whose causal influence on outputs can be measured and amplified via differential steering to achieve interpretable personalization.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Preference Heads in Large Language Models: A Mechanistic Framework for Interpretable Personalization
LLMs contain identifiable Preference Heads whose causal influence on outputs can be measured and amplified via differential steering to achieve interpretable personalization.