AI Paper Atlas

Research map / Large language models / Alignment and preferences

Circular-DPO: Aligning Multi-Stage 3D Generative Models via Preference Feedback Loop

Zejian Li, Jiarui Ma, Han Xu, Weiting Zheng, Yangrui Zhu, Chenye Meng, Pei Chen, Ling Yang, Zhiyuan Yang, Changyuan Yang, Guang Yang, Immanuel Koh and 1 more

CVPR 2026 · Large language models › Alignment and preferences · approach: DPO design choices

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 the data feedback loop that trains stage 1 on pairs from the aligned stage 2, versus independent stage-wise DPO (ablation)Circular-DPO on Trellis · Test set of 100 promptsImageReward: -0.3663 -> -0.3146Improved
add the data feedback loop (versus independent stage-wise DPO) (ablation)Circular-DPO on Trellis · Test set of 100 promptsReward3D: -0.5969 -> -0.5607Improved
add quality-gap filter and weight w1 (ablation)Circular-DPO on Trellis · Test set of 100 promptsReward3D: -0.6199 -> -0.5607Improved
add quality-gap filter and weight w1 (ablation)Circular-DPO on Trellis · Test set of 100 promptsImageReward: -0.3246 -> -0.3146Improved
add preference reliability weight w2 (ablation)Circular-DPO on Trellis · Test set of 100 promptsImageReward: -0.4908 -> -0.3146Improved
use DPO instead of plain fine-tuning on stage 1 (ablation)Trellis sparse structure model · Test set of 100 promptsImageReward: -0.4187 -> -0.3748Improved
use DPO instead of plain fine-tuning on stage 2 (ablation)Trellis structured latent model · Test set of 100 promptsImageReward: -0.4410 -> -0.5512Hurt
apply DPO only to stage 2 (without stage 1 loop), vs untuned Trellis (ablation)Trellis · Test set of 100 promptsReward3D: -0.7137 -> -0.7112No clear effect
apply Circular-DPO to TrellisTrellis · Test set of 100 promptsImageReward: -0.4596 -> -0.3146Improved
apply Circular-DPO to MVDreamMVDream · Test set of 100 promptsImageReward: -0.3948 -> -0.3747Improved
compare Circular-DPO on Trellis against score-distillation DreamRewardCircular-DPO on Trellis · Test set of 100 promptsReward3D: -0.5607 vs 0.2733Hurt

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