Skywork-VL Reward: An Effective Reward Model for Multimodal
  Understanding and Reasoning

Xiaokun Wang; Chris; Jiangbo Pei; Wei Shen; Yi Peng; Yunzhuo Hao; Weijie Qiu; Ai Jian; Tianyidan Xie; Xuchen Song; Yang Liu; Yahui Zhou

Paper · arXiv 2505.07263

Skywork-VL Reward: An Effective Reward Model for Multimodal Understanding and Reasoning

Xiaokun Wang, Chris, Jiangbo Pei, Wei Shen, Yi Peng, Yunzhuo Hao, Weijie Qiu, Ai Jian, Tianyidan Xie, Xuchen Song, Yang Liu, Yahui Zhou

30 upvotesMay 12, 2025arXiv 预印本

AI 摘要

Skywork-VL Reward is a multimodal reward model that uses a large-scale preference dataset and Qwen2.5-VL-7B-Instruct architecture to achieve state-of-the-art performance in multimodal reasoning.

multimodal reward modelmultimodal understandingQwen2.5-VL-7B-Instructreward headmulti-stage fine-tuningpairwise ranking lossmultimodal VL-RewardBenchtext-only RewardBenchMixed Preference Optimizationmultimodal reasoning capabilitiesgeneral-purpose reward modelsmultimodal alignment

Abstract

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.