Saved in:
Bibliographic Details
Main Authors: Wang, Xiaokun, Wang, Peiyu, Pei, Jiangbo, Shen, Wei, Peng, Yi, Hao, Yunzhuo, Qiu, Weijie, Jian, Ai, Xie, Tianyidan, Song, Xuchen, Liu, Yang, Zhou, Yahui
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.07263
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • We propose Skywork-VL Reward, a multimodal reward model that provides reward signals for both multimodal understanding and reasoning tasks. Our technical approach comprises two key components: First, we construct a large-scale multimodal preference dataset that covers a wide range of tasks and scenarios, with responses collected from both standard vision-language models (VLMs) and advanced VLM reasoners. Second, we design a reward model architecture based on Qwen2.5-VL-7B-Instruct, integrating a reward head and applying multi-stage fine-tuning using pairwise ranking loss on pairwise preference data. Experimental evaluations show that Skywork-VL Reward achieves state-of-the-art results on multimodal VL-RewardBench and exhibits competitive performance on the text-only RewardBench benchmark. Furthermore, preference data constructed based on our Skywork-VL Reward proves highly effective for training Mixed Preference Optimization (MPO), leading to significant improvements in multimodal reasoning capabilities. Our results underscore Skywork-VL Reward as a significant advancement toward general-purpose, reliable reward models for multimodal alignment. Our model has been publicly released to promote transparency and reproducibility.