AIApr 21

DT2IT-MRM: Debiased Preference Construction and Iterative Training for Multimodal Reward Modeling

arXiv:2604.1954494.6Has Code
AI Analysis

This work addresses noise and biases in multimodal preference data for training reward models, which is critical for aligning multimodal LLMs with human preferences.

DT2IT-MRM introduces a debiased preference construction pipeline and iterative training framework to improve multimodal reward models, achieving state-of-the-art performance on three major benchmarks.

Multimodal reward models (MRMs) play a crucial role in aligning Multimodal Large Language Models (MLLMs) with human preferences. Training a good MRM requires high-quality multimodal preference data. However, existing preference datasets face three key challenges: lack of granularity in preference strength, textual style bias, and unreliable preference signals. Besides, existing open-source multimodal preference datasets suffer from substantial noise, yet there is a lack of effective and scalable curation methods to enhance their quality. To address these limitations, we propose \textbf{DT2IT-MRM}, which integrates a \textbf{D}ebiased preference construction pipeline, a novel reformulation of text-to-image (\textbf{T2I}) preference data, and an \textbf{I}terative \textbf{T}raining framework that curates existing multimodal preference datasets for \textbf{M}ultimodal \textbf{R}eward \textbf{M}odeling. Our experimental results show that DT2IT-MRM achieves new \textbf{state-of-the-art} overall performance on three major benchmarks: VL-RewardBench, Multimodal RewardBench, and MM-RLHF-RewardBench.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes