paper-with-me

홈 › Papers

DeDPO: Debiased Direct Preference Optimization for Diffusion Models

2026-02-05 · Khiem Pham, Quang Nguyen, Tung Nguyen, Jingsen Zhu, Michele Santacatterina, Dimitris Metaxas, Ramin Zabih arxiv

Direct Preference Optimization (DPO) has emerged as a predominant alignment method for diffusion models, facilitating off-policy training without explicit reward modeling. However, its reliance on large-scale, high-quality human preference labels presents a severe cost and scalability bottleneck. To overcome this, We propose a semi-supervised framework augmenting limited human data with a large corpus of unlabeled pairs annotated via cost-effective synthetic AI feedback. Our paper introduces Debiased DPO (DeDPO), which uniquely integrates a debiased estimation technique from causal inference into the DPO objective. By explicitly identifying and correcting the systematic bias and noise inherent in synthetic annotators, DeDPO ensures robust learning from imperfect feedback sources, including self-training and Vision-Language Models (VLMs). Experiments demonstrate that DeDPO is robust to the variations in synthetic labeling methods, achieving performance that matches and occasionally exceeds the theoretical upper bound of models trained on fully human-labeled data. This establishes DeDPO as a scalable solution for human-AI alignment using inexpensive synthetic supervision.

📄 PDF Abstract BibTeX arXiv:2602.06195

Code (0)

등록된 구현이 없습니다.

Tasks

Causal Inference

Similar Papers 제목 키워드 기반

BideDPO: Conditional Image Generation with Simultaneous Text and Condition Alignment

2025-11-24 · Dewei Zhou, Mingwei Li, Zongxin Yang, Yu Lu 외 arxiv

Conditional image generation enhances text-to-image synthesis with structural, spatial, or stylistic priors, but current methods face challenges in handling conflicts between sources. These include 1) input-level conflic…

Conditional Image Generation

A Statistical Framework for Alignment with Biased AI Feedback

2026-02-09 · Xintao Xia, Zhiqiu Xia, Linjun Zhang, Zhanrui Cai arxiv

Modern alignment pipelines are increasingly replacing expensive human preference labels with evaluations from large language models (LLM-as-Judge). However, AI labels can be systematically biased compared to high-quality…

Computational Efficiency

Forward KL Regularized Preference Optimization for Aligning Diffusion Policies

2024-09-09 · Zhao Shan, Chenyou Fan, Shuang Qiu, Jiyuan Shi 외

Diffusion models have achieved remarkable success in sequential decision-making by leveraging the highly expressive model capabilities in policy learning. A central problem for learning diffusion policies is to align the…

D4RLDecision MakingReinforcement Learning (RL)Sequential Decision Making

Refining Alignment Framework for Diffusion Models with Intermediate-Step Preference Ranking

2025-02-01 · Jie Ren, Yuhang Zhang, Dongrui Liu, Xiaopeng Zhang 외

Direct preference optimization (DPO) has shown success in aligning diffusion models with human preference. Previous approaches typically assume a consistent preference label between final generations and noisy samples at…

Debiasing Multimodal Large Language Models via Noise-Aware Preference Optimization

2025-03-23 · CVPR 2025 1 · Zefeng Zhang, Hengzhu Tang, Jiawei Sheng, Zhenyu Zhang 외

Multimodal Large Language Models excel in various tasks, yet often struggle with modality bias, where the model tends to rely heavily on a single modality and overlook critical information in other modalities, which lead…