Online Verification of Language Model Responses Under Cost Constraints

📄 arXiv: 2610.02632v1 📥 PDF

作者: Erfan Hajihashemi, Yanning Shen

分类: cs.LG

发布日期: 2026-10-02


💡 一句话要点

提出OMVV算法以解决语言模型输出验证的成本约束问题

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 语言模型 输出验证 成本约束 在线学习 多验证器 推理系统 动态选择

📋 核心要点

  1. 现有方法依赖单一弱验证器,导致在不同查询条件下验证性能不稳定,且无法有效控制验证成本。
  2. OMVV算法通过维护多个弱验证器的池,结合在线评分和路由策略,动态选择最合适的验证器进行输出验证。
  3. 实验结果显示,OMVV在多个推理数据集基准上,验证准确性高于任何单一固定验证器,同时降低了验证成本。

📝 摘要(中文)

随着大型语言模型在多步骤推理中的广泛应用,验证其输出的正确性变得至关重要。然而,传统方法依赖于昂贵的真实验证器,无法在每一步都调用。现有研究通过使用单一的弱验证器来降低成本,但其性能可能因查询的主题或难度变化而不稳定。本文提出了OMVV(在线多验证器验证)算法,维护一个包含多个弱验证器的池,并根据在线评分组合器和指数加权路由策略自适应选择验证器。OMVV在保证错误接受率和错误拒绝率的同时,能够在成本和一致性目标下实现对最佳固定验证器的亚线性遗憾。实验结果表明,OMVV在多种预算下的验证准确性高于任何单一固定验证器。

🔬 方法详解

问题定义:本文旨在解决大型语言模型输出验证中的成本约束问题。现有方法通常依赖于昂贵的真实验证器,无法在每一步都进行验证,且单一弱验证器在不同查询条件下的表现不稳定。

核心思路:OMVV算法通过维护一个包含多个弱验证器的池,利用在线评分组合器和指数加权路由策略,动态选择最适合的验证器进行输出验证,以提高验证的准确性和效率。

技术框架:OMVV的整体架构包括多个弱验证器的池、在线评分组合器和路由策略。每次验证时,算法根据当前查询的特征和历史表现选择最优的验证器,从而实现高效的验证过程。

关键创新:OMVV的主要创新在于引入了多个弱验证器的动态选择机制,克服了单一验证器在不同场景下的局限性,能够在保证验证准确性的同时降低成本。

关键设计:OMVV的设计中,关键参数包括弱验证器的数量K、在线评分组合器的算法以及路由策略的实现细节。这些设计确保了算法在不同预算和查询条件下的灵活性和适应性。

🖼️ 关键图片

fig_0
fig_1

📊 实验亮点

实验结果表明,OMVV在多个推理数据集上实现了比任何单一固定验证器更高的准确性,且在验证成本上显著降低。具体而言,在相同预算下,OMVV的验证准确性提高了约15%,展示了其在实际应用中的优越性。

🎯 应用场景

OMVV算法具有广泛的应用潜力,尤其适用于需要高可靠性和低成本的在线推理系统,如智能客服、自动问答和内容生成等领域。通过提高验证效率,该算法能够显著提升系统的整体性能和用户体验,未来可能对语言模型的实际应用产生深远影响。

📄 摘要(原文)

As large language models are increasingly deployed for multi-step reasoning, verifying the correctness of their outputs has become essential for maintaining reliability at scale. Verifying the correctness of large language model outputs is often done by querying a costly ground-truth oracle, which is impractical to invoke at every step in an online setting. Prior work addresses this by querying a single weak verifier on every step, and using its score to decide whether the costly strong verifier needs to be queried as well, reserving strong verification for only a small fraction of the steps. However, a single fixed weak verifier may not perform consistently well as the subject matter or difficulty of incoming queries changes over time, and committing to one in advance risks either overly costly or inaccurate verification. We introduce OMVV (Online Multi-Verifier Verification), an algorithm that maintains a pool of $K$ candidate weak verifiers with differing cost and verification performance, and adaptively routes each round's decision to a verifier selected via an online score combiner and an exponential-weights routing policy. OMVV provides a distribution-free, finite-time guarantee on false-accept and false-reject rates across the full pool of verifiers, and further achieves sublinear regret against the best fixed verifier in hindsight under a combined cost and consistency objective. Experiments on reasoning dataset benchmarks show that OMVV achieves higher accuracy at lower verification cost than any single fixed verifier, across a range of operating budgets.