AI Paper Atlas

Research map / Large language models / Alignment and preferences

Multi-objective Large Language Model Alignment with Hierarchical Experts

Zhuo Li, Guodong DU, Weiyang Guo, Yigeng Zhou, Xiucheng Li, Wenya Wang, Fangming Liu, Yequan Wang, Deheng Ye, Min Zhang, Jing Li

ICLR 2026 · Large language models › Alignment and preferences · approach: Robust and calibrated DPO · 10 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
use HoE instead of MODLlama3.1-8B · Psoups 3-objective, Helpfulreward score: 0.94 -> 3.05Improved
use HoE instead of PADLlama3.1-8B · Psoups 3-objective, HHHreward score: 1.06 vs 1.03No clear effect
add one router expert alone to two single-objective LoRA experts (ablation)HoE · Pareto frontier, two objectivesPareto frontier coverageImproved
add one LoRA expert to the two fixed experts (ablation)HoE · Pareto frontier, two objectivesPareto frontier coverageMixed
add one LoRA expert and one router expert (ablation)HoE · Pareto frontier, two objectivesPareto frontier coverageImproved
add a second LoRA expert (2 LoRA) (ablation)HoE · Pareto frontier, two objectivesPareto frontier coverageImproved
lower LoRA rank below 128 (ablation)HoE · Math taskperformanceHurt
use Tchebycheff instead of linear scalarization (ablation)router training · Pareto frontier stabilitytraining stabilityImproved
use HoE instead of MOD at inferenceLlama2-7B · 3-objective alignmentrelative inference cost: 3.10 -> 1.23Improved
use HoE multi-objective LoRA expert instead of PCB/FR/TIES mergingChinese/Math/Code LLaMA · GSM8Kaccuracy: 56.2 (TIES) vs 50.4Hurt
rescale task-SVD LoRA with factor 1.9 (ablation)LLaMA2-7B-Chat Math LoRA · GSM8Kaccuracy: 26% -> 68%Improved

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