AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Dual Mechanisms of Value Expression: Intrinsic vs. Prompted Values in Large Language Models

Jongwook Han, Jongwon Lim, Injin Kong, Yohan Jo

ICML 2026 · Large language models › Language, safety and interpretability · approach: Mechanistic analysis of attention and representations · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
steer along intrinsic or prompted value vectorsQwen2.5-7B-Instruct · PVQ questionnaire (avg over 5 languages)mean score delta: +1.74 (intrinsic), +2.21 (prompted)Improved
remove the prompted component from the intrinsic vector (orthogonalize) (ablation)intrinsic value vector · PVQ questionnaire (avg)mean score delta: +1.74 -> +0.47Hurt
remove the intrinsic component from the prompted vector (orthogonalize) (ablation)prompted value vector · PVQ questionnaire (avg)mean score delta: +2.21 -> +1.62Mixed
orthogonalize intrinsic vector, free-form format (ablation)intrinsic value vector · PVQ free-form (avg)mean score delta: +0.98 -> +0.48Hurt
apply English-extracted vectors to other languagesvalue vectors · multilingual PVQ and dilemmassteering effectNo clear effect
scale activations of shared neurons versus unique neuronsvalue neurons · value expressionvalue expression increaseImproved
use intrinsic steering instead of prompted steeringvalue vectors · situational dilemmas diversity (Achievement example)Distinct-2: 0.342 -> 0.362Improved
orthogonalize prompted vector against intrinsic (ablation)prompted value vector · situational dilemmas diversity (Achievement example)Distinct-2: 0.342 -> 0.203Hurt
orthogonalize intrinsic vector against prompted (ablation)intrinsic value vector · situational dilemmas diversity (Achievement example)Distinct-2: 0.362 -> 0.402Improved
steer along the mean delta (prompted minus intrinsic) directionLlama-3.1-8B-Instruct with compliance system prompt · AdvBenchASR@9: 13.3% -> 97.2%Improved
steer along the mean delta directionQwen2.5-7B-Instruct with compliance system prompt · HarmBenchASR@9: 52.4% -> 83.0%Improved
steer along compliance direction on non-value instruction tasksQwen2.5-7B-Instruct · gender-specific translation vs strict JSON formattingcomplianceMixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy