AI Paper Atlas

Research map / Large language models / Alignment and preferences

OrthAlign: Orthogonal Subspace Decomposition for Non-Interfering Multi-Objective Alignment

Liang Lin, Zhihao Xu, Junhao Dong, Jian Zhao, Yuchen Yuan, Guibin Zhang, Miao Yu, Yiming Zhang, Zhengtao Yao, Huahui Yi, HAICHUAN TANG, Dongrui Liu and 2 more

ICLR 2026 · Large language models › Alignment and preferences · approach: Robust and calibrated DPO

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
orthogonal subspace updates (OrthAlign) instead of strongest baseline on 3-objective alignmentOrthAlign · HelpSteer2 / UltraFeedback, Llama3-SFTaverage score: 70.13 -> 75.95 (HelpSteer2 vs Knots)Improved
OrthAlign vs RSDPO on helpful win rateOrthAlign · UltraFeedback Llama3-SFThelpful win rate: 71.92 -> 71.57No clear effect
OrthAlign vs DPO on truthfulnessOrthAlign · UltraFeedback Llama3-SFTTruthfulQA MC2: 67.03 -> 66.58No clear effect
add OrthAlign subspace projection to DPO (ablation)DPO · HelpSteer2+SafeRLHFharmless rate: 71.24 -> 93.84Improved
add OrthAlign subspace projection to MODPO (ablation)MODPO · HelpSteer2+SafeRLHFharmless rate: 48.46 -> 79.32Improved
add OrthAlign subspace projection to SPO (ablation)SPO · HelpSteer2+SafeRLHFhelpful win rate: 61.24 -> 67.28Improved
increase fixed subspace rank from 12 to 26 (ablation)OrthAlign · Helpsteer2+SafeRLHF, Llama3-SFTharmless rate: 93.80 -> 81.34Hurt
vary subspace rank (14 onward) (ablation)OrthAlign · Helpsteer2+SafeRLHF, Llama3-SFThelpful win rate: 63.59 ~ 65.79No clear effect
reverse training order to Helpful->Harmless (ablation)OrthAlign · UltraFeedback+SafeRLHFaverage score: 79.59 -> 77.27Mixed

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