AI Paper Atlas

Research map / Large language models / Alignment and preferences

Multi-dimensional Preference Alignment by Conditioning Reward Itself

JIHO JANG, Jin-Young Kim, Kyungjune Baek, Nojun Kwak

ECCV 2026 · Large language models › Alignment and preferences · approach: DPO design choices · 3 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
add dimensional reward dropout (vs rate 0.0) (ablation)MCDPO · Pick-a-Pic v2 test (SD1.5)CLIP win rate: 49.2 -> 57.6Improved
add dimensional reward dropout (vs rate 0.0) (ablation)MCDPO · Pick-a-Pic v2 test (SD1.5)Avg win rate: 80.1 -> 81.1Improved
add MCSFT pre-training (ablation)MCDPO · Pick-a-Pic v2 test (SD1.5)Avg win rate: 60.5 -> 81.1Improved
add context-aware reward conditioning module (ablation)MCDPO · Pick-a-Pic v2 test (SD1.5)Avg win rate: 70.1 -> 81.1Improved
use MCDPO alignment after MCSFT instead of full MCSFT training aloneMCSFT · Pick-a-Pic v2 test (SD1.5)Avg win rate: 73.2 -> 81.1Improved
train MCDPO on conflict-only pairs instead of random pairsMCDPO · Pick-a-Pic v2 test (SD1.5)Avg win rate: 81.1 -> 77.0Hurt
train standard DPO on conflict-only pairs instead of random pairsDPO · Pick-a-Pic v2 test (SD1.5)Avg win rate: 66.7 -> 59.0Hurt
use MCDPO on random pairs instead of filtered conflict-free DPOFiltered DPO · Pick-a-Pic v2 test (SD1.5)Avg win rate: 78.7 -> 81.1Improved
use MCDPO instead of standard DPODPO · Pick-a-Pic v2 test (SD1.5)Avg win rate: 66.7 -> 81.1Improved
use conditional multi-dimension MCDPO instead of Aesthetic-only Conditional DPO specialistCond. DPO specialist · Pick-a-Pic v2 test (SD1.5)Aesthetic win rate: 88.1 -> 91.8Improved
amplify CLIP only at test timeMCDPO · Pick-a-Pic v2 test (SD1.5)CLIP win rate: 57.6 -> 59.0Improved
add Aes-CLIP temporal scheduling at test timeMCDPO · Pick-a-Pic v2 test (SD1.5)CLIP win rate: 57.6 -> 59.6Improved

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