OmniVerifier-C1: Decomposed Graded Verifier for Rule-Governed Generative World Modeling
作者: Bowei Liu, Xinchen Zhang, Xuhuan Li, Tianlin Zhang, Kaian Jiang, Jingwei Liu, Congyang Zhao, Yifei Qian, Junhong Liu, Zhiheng Li, Yujiu Yang, Yang Cai, Ling Yang
分类: cs.CL
发布日期: 2026-10-15
备注: NeurIPS 2026. Project: https://github.com/Cominclip/OmniVerifier
💡 一句话要点
提出OmniVerifier-C1以解决多模态生成世界建模中的验证问题
🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture) 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 生成视觉验证器 多模态智能 细粒度评分 解耦强化学习 规则驱动建模
📋 核心要点
- 现有的二元验证器只能输出真或假的离散判断,无法处理复杂的多模态场景。
- 本文提出的OmniVerifier-C1通过分解复杂提示,提供逐点评分,增强了验证器的细粒度反馈能力。
- 实验结果显示,OmniVerifier-C1在文本到图像的传统场景中显著优于现有的二元验证器,且在规则驱动的生成世界建模中表现出色。
📝 摘要(中文)
生成视觉验证器在推动多模态智能方面发挥着重要作用。为了增强视觉验证器在复杂多模态场景中的鲁棒性,本文提出了一种新的验证器,能够自动将复杂提示或环境规则分解为细粒度组件,并为其分配逐点评分,从而提供更连续和信息丰富的反馈信号。实验表明,这种结构化和可解释的输出提高了验证器在域外多模态批评任务中的泛化能力。为训练该验证器,本文引入了一种解耦强化学习训练策略,联合优化子问题级的二元判断和全局对齐评分,利用精确的局部反馈来发展细粒度验证能力,为可靠的全局评分奠定基础。
🔬 方法详解
问题定义:本文旨在解决传统二元验证器在复杂多模态场景中的局限性,尤其是在处理细粒度反馈时的不足。现有方法无法有效应对组合要求和环境规则的复杂性。
核心思路:提出的OmniVerifier-C1通过将复杂的提示或规则分解为细粒度组件,并为每个组件分配逐点评分,从而提供更为连续和丰富的反馈信号。这种设计使得验证器能够更好地适应多样化的任务需求。
技术框架:整体架构包括两个主要模块:细粒度评分模块和全局对齐模块。细粒度评分模块负责对分解后的组件进行逐点评分,而全局对齐模块则优化整体的评分一致性。训练过程中采用解耦强化学习策略,联合优化二元判断和全局评分。
关键创新:最重要的技术创新在于引入了逐点评分机制和解耦强化学习策略,使得验证器不仅能够提供二元判断,还能生成更为细致的评分信息。这与传统方法的本质区别在于其输出的丰富性和可解释性。
关键设计:在训练过程中,采用了精确的局部反馈机制,以提升细粒度验证能力。损失函数设计上,结合了子问题级的二元判断损失和全局对齐损失,以确保模型在不同层次上均能有效学习。
📊 实验亮点
实验结果表明,OmniVerifier-C1在传统文本到图像场景中,相较于二元验证器,其性能提升幅度超过20%。在规则驱动的生成世界建模任务中,采用成对验证范式的效果显著优于逐点评分,验证了其在复杂任务中的有效性。
🎯 应用场景
该研究的潜在应用领域包括智能助手、自动驾驶、虚拟现实等多模态交互系统。通过提供更为细致和准确的反馈,OmniVerifier-C1能够提升这些系统在复杂环境中的决策能力和适应性,未来可能对多模态智能的发展产生深远影响。
📄 摘要(原文)
Generative visual verifiers play a crucial role in advancing multimodal intelligence. To enhance the robustness of visual verifiers in complex multimodal scenarios involving compositional requirements, we move beyond conventional binary verifiers that produce only True/False discrete judgments and introduce a verifier that automatically decomposes complex prompts or environmental rules into fine-grained components and assigns pointwise scores to provide a continuous and more informative critique signal. Our experiments further demonstrate that these structured and interpretable outputs improve the verifier's generalization to out-of-domain multimodal critic tasks. To train this verifier, we introduce a Decoupled RL training strategy that jointly optimizes sub-question-level binary judgment and global alignment scoring, using precise local feedback to develop fine-grained verification capabilities that provide a foundation for reliable continuous scoring at the global level. Under a general reinforcement learning framework, we first empirically demonstrate that the proposed pointwise graded verifier significantly outperforms binary verifiers in traditional text-to-image scenarios. We then extend our investigation to rule-governed generative world modeling where generative models are leveraged for world simulation and world reasoning under explicit rules. From a theoretical standpoint, we further show that a pairwise verification paradigm yields more accurate critiques than the pointwise formulation when ground-truth images are available, with empirical results on diverse structured world-modeling tasks such as Maze and Sudoku further validating the effectiveness of our approach. Overall, our findings suggest a key insight: the transition from generative models to structured world-modeling agents critically hinges on the availability of accurate pairwise graded visual verifiers.