Boosting the Power of Small Multimodal Reasoning Models to Match Larger Models with Self-Consistency Training

📄 arXiv: 2311.14109v2 📥 PDF

作者: Cheng Tan, Jingxuan Wei, Zhangyang Gao, Linzhuang Sun, Siyuan Li, Ruifeng Guo, Bihui Yu, Stan Z. Li

分类: cs.AI

发布日期: 2023-11-23 (更新: 2024-07-03)

备注: Accepted by ECCV 2024

🔗 代码/项目: GITHUB

💡 一句话要点

提出MC-CoT以提升小型多模态推理模型的性能

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

关键词: 多模态推理 自一致性训练 推理依据生成 模型性能提升 智能问答 图像理解 投票机制

📋 核心要点

  1. 现有的多模态推理方法在生成推理依据的质量上存在不足，影响了模型的整体性能。
  2. 本文提出了一种名为MC-CoT的自一致性训练策略，通过生成多个推理依据并进行投票选择，提升推理依据的质量。
  3. 实验结果显示，采用该方法后，模型在多个基准测试上性能显著提升，小型模型的表现可与大型模型相媲美。

📝 摘要（中文）

多模态推理是一项复杂的任务，需要模型跨多种模态进行推理以回答问题。现有方法在将语言和视觉模态结合的两阶段推理框架中取得了一定进展，但生成的推理依据质量不足，导致模型性能不佳。本文强调推理依据在模型推理中的重要性，提出了一种自一致性训练策略MC-CoT，通过生成多个推理依据和答案，并通过投票选择最准确的结果，从而提高生成推理依据的质量和答案的准确性。实验表明，该方法显著提升了模型在多个基准测试中的表现，甚至小型基础模型也能达到与大型模型相当的效果，展示了推理依据在多模态推理中的潜力。

🔬 方法详解

问题定义：本文旨在解决多模态推理中生成推理依据质量不足的问题。现有方法通常将推理依据生成与答案推断分开，导致推理依据的准确性不足，从而影响最终答案的质量。

核心思路：论文提出的MC-CoT策略通过生成多个推理依据和答案，并利用投票机制选择最优结果，旨在提高推理依据的质量和模型的整体推理能力。

技术框架：整体架构包括两个主要阶段：首先生成多个推理依据和答案，然后通过投票机制评估和选择最准确的推理依据。该框架强调了推理依据生成的多样性和准确性。

关键创新：MC-CoT的核心创新在于自一致性训练策略，通过生成多个候选推理依据并进行投票选择，显著提高了推理依据的质量，与传统方法相比，能够更有效地利用推理依据提升模型性能。

关键设计：在实现过程中，设计了特定的损失函数以优化推理依据的生成质量，并采用了多模态融合技术来增强模型对不同模态信息的理解能力。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果表明，采用MC-CoT策略后，模型在多个基准测试中的性能提升显著，尤其是小型模型的表现可与大型模型相媲美，提升幅度达到20%以上，展示了该方法在多模态推理中的有效性。

🎯 应用场景

该研究的潜在应用领域包括智能问答系统、图像理解、视频分析等多模态交互场景。通过提升小型模型的推理能力，能够在资源受限的环境中实现高效的多模态推理，具有重要的实际价值和广泛的应用前景。

📄 摘要（原文）

Multimodal reasoning is a challenging task that requires models to reason across multiple modalities to answer questions. Existing approaches have made progress by incorporating language and visual modalities into a two-stage reasoning framework, separating rationale generation from answer inference. However, these approaches often fall short due to the inadequate quality of the generated rationales. In this work, we delve into the importance of rationales in model reasoning. We observe that when rationales are completely accurate, the model's accuracy significantly improves, highlighting the need for high-quality rationale generation. Motivated by this, we propose MC-CoT, a self-consistency training strategy that generates multiple rationales and answers, subsequently selecting the most accurate through a voting process. This approach not only enhances the quality of generated rationales but also leads to more accurate and robust answers. Through extensive experiments, we demonstrate that our approach significantly improves model performance across various benchmarks. Remarkably, we show that even smaller base models, when equipped with our proposed approach, can achieve results comparable to those of larger models, illustrating the potential of our approach in harnessing the power of rationales for improved multimodal reasoning. The code is available at https://github.com/chengtan9907/mc-cot.