AI Paper Atlas

Research map / Large language models / Alignment and preferences

The Sign Estimator: Preference Modeling for LLM Alignment under Heterogeneity

Ali Aouad, Aymane El Gadarri, Vivek Farias

ICML 2026 · Large language models › Alignment and preferences · approach: Reward modeling for diverse preferences

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
replace cross-entropy loss with Sign Estimator binary classification lossRLHF reward model · SHPaccuracy: 66.50 -> 68.33Improved
replace cross-entropy loss with Sign Estimator lossRLHF reward model · HH-Anthropicaccuracy: 66.89 -> 68.58Improved
replace cross-entropy loss with Sign Estimator lossRLHF reward model · CoValaccuracy: 78.09% to 80.72%Improved
replace cross-entropy loss with Sign Estimator lossRLHF reward model · CoVal (trained on HH-Anthropic)accuracy: 69.94% to 73.07%Improved
replace cross-entropy loss with Sign Estimator lossRLHF estimator · semi-synthetic digital twinsangular error to mean utility: 63 -> 41 degrees (about 35%)Improved
replace cross-entropy loss with Sign Estimator lossRLHF estimator · semi-synthetic digital twinsdisagreement rate: roughly 12% to 8%Improved
use EM with K=2 segments instead of pooled RLHF (ablation)RLHF · semi-synthetic digital twinsangle error, disagreement, accuracyImproved
increase EM segments from K=2 to K=3 or 5 (ablation)EM reward learning · semi-synthetic digital twinsaccuracy and error metricsHurt
use EM with K=3 segments instead of pooled RLHF (ablation)RLHF · SHPaccuracy: 66.50 -> 60.83Hurt
increase heterogeneity scale from 1x to 12x (ablation)Sign vs RLHF · semi-synthetic digital twinsrelative disagreement reduction delta%: 29.22% -> 38.49%Improved
increase heterogeneity scale 1x to 12x (ablation)RLHF · semi-synthetic digital twinsangle error: 63.24 -> 74.36Hurt
increase heterogeneity scale 1x to 12x (ablation)Sign Estimator · semi-synthetic digital twinsangle error: 41.55 -> 49.68Hurt

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