AI Paper Atlas

Research map / Large language models / Alignment and preferences

Generative Value Conflicts Reveal LLM Priorities

Andy Liu, Kshitish Ghate, Mona T. Diab, Daniel Fried, Atoosa Kasirzadeh, Max Kleiman-Weiner

ICLR 2026 · Large language models › Alignment and preferences · approach: Steering and correcting LLM behavior

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
filter scenarios with LLM judge (ablation)ConflictScope · HHH scenariosobserved agreement (lower is better): 0.824 -> 0.786Improved
filter scenarios with LLM judge (ablation)ConflictScope · HHH scenariosLikert difference rate: 0.818 -> 0.801No clear effect
use two-stage generation instead of single-stage (ablation)ConflictScope · HHH scenariosobserved agreement (lower is better): 0.898 -> 0.824 (unfiltered)Improved
use values and prompt templates instead of direct generation (ablation)ConflictScope · HHH scenariosobserved agreement (lower is better): 0.852 -> 0.824 (unfiltered)Improved
two-stage generation vs single-stage (ablation)ConflictScope · HHH scenariosLikert difference rate: 0.854 (single-stage) -> 0.818 (unfiltered two-stage)Hurt
switch from multiple-choice to open-ended evaluation14 target LLMs · Personal-Protectiveaverage rank of protective values: 1.7 -> 4.5Hurt
add system prompt describing target value ranking14 target LLMs · 3 value sets, open-endednormalized alignment effect: average effect 0.145Improved
add steering system promptClaude 3.5 Haiku · 3 value setsnormalized alignment effect: 0.01No clear effect
steering prompt on ModelSpec vs HHH and Personal-Protective14 target LLMs · ModelSpecnormalized alignment effectMixed

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