Decoding Affective Nuances: Enhancing MLLMs via Hierarchical Emotion Reasoning and Contrastive Discriminative Pruning

📄 arXiv: 2609.36782v1 📥 PDF

作者: Cheng Ye, Weidong Chen, Zhaobo Qi, Beier Zhu, Zhendong Mao

分类: cs.CV

发布日期: 2026-09-29

备注: 19 pages, 8 figures

💡 一句话要点

提出DAN框架以解决多模态大语言模型情感推理不足问题

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

关键词: 多模态大语言模型 情感推理 层次情感推理 对比性区分 视觉修剪 情感分析 智能客服

📋 核心要点

  1. 现有的多模态大语言模型在情感推理方面存在不足，难以准确区分细微情感差异。
  2. 本文提出DAN框架，通过层次情感推理链和对比性区分视觉修剪来解决情感推理中的归因不足和区分不足问题。
  3. 在多个基准测试中，DAN显著提升了情感细微差别的区分能力，尤其在WebEmo25数据集上取得了显著的性能提升。

📝 摘要（中文）

尽管多模态大语言模型（MLLMs）在客观理解任务中表现出色，但在情感推理方面的表现仍显著低于人类标准。我们认为其主要原因在于MLLMs难以基于细粒度视觉证据可靠地区分语义相近的情感。为了解决这一问题，本文提出了一种无训练的推理优化框架DAN，包含层次情感推理链（HERC）和对比性区分视觉修剪（CDVP），显著提高了情感细微差别的区分能力，尤其在WebEmo25数据集上实现了+10.47%的性能提升。

🔬 方法详解

问题定义：本文旨在解决多模态大语言模型在情感推理中对细粒度情感线索的不足归因和区分能力差的问题。现有方法往往无法有效识别语义相近情感之间的细微差异，导致情感判断模糊不清。

核心思路：提出一种无训练的推理优化框架DAN，通过层次情感推理链（HERC）来增强细粒度情感线索的归因能力，并设计对比性区分视觉修剪（CDVP）来区分语义相近的情感。

技术框架：DAN框架主要包括两个模块：HERC用于整合场景和对象级别的细粒度线索，CDVP用于隔离具有区分性的视觉标记。整体流程为：首先通过HERC进行情感推理，然后利用CDVP进行视觉标记的对比分析。

关键创新：HERC和CDVP是本文的核心创新，HERC通过软门控推理提升了情感线索的归因能力，而CDVP则通过计算相似情感的注意力分布绝对差异来实现情感区分，这与现有方法的全局推理方式有本质区别。

关键设计：在HERC中，采用了软门控机制以平衡不同层次的情感线索，而在CDVP中，通过设计特定的损失函数来优化视觉标记的选择，确保最终情感类别的准确性。整体网络结构经过精心设计，以支持高效的推理过程。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

在多个基准测试中，DAN框架显著提升了情感细微差别的区分能力，特别是在WebEmo25数据集上，使用Qwen3-VL-8B-Instruct模型实现了+10.47%的性能提升，展示了其在情感推理中的有效性。

🎯 应用场景

该研究的潜在应用领域包括情感分析、社交媒体监测、智能客服等，能够帮助系统更准确地理解和响应用户情感，提高人机交互的自然性和有效性。未来，该框架可能在情感计算和情感智能领域产生深远影响。

📄 摘要（原文）

While multimodal large language models (MLLMs) have demonstrated exceptional capabilities in objective understanding tasks, their performance in affective reasoning still falls significantly short of human standards. We attribute it to a central capability gap: MLLMs are difficult to reliably distinguish semantically proximal emotions based on fine-grained visual evidence, which could be decoupled as two limitations: 1) Insufficient Attribution. The global reasoning paradigm of conventional MLLMs severely dilutes fine-grained emotion cues, where subtle emotional states are usually implicitly encoded, thereby generating emotional misjudgments in complex scenarios. 2) Insufficient Discrimination. Existing methods could only identify regions generally associated with emotions, which fails to distinguish discriminative regions between semantically similar emotions, leading to ambiguous emotion judgements. To overcome these limitations, we present a training-free inference-time optimization framework, named Decoding Affective Nuances (DAN). Specifically, we propose a Hierarchical Emotional Reasoning Chain (HERC) that enhances the insufficient attribution by harmonizing fine-grained scene/object-level cues and performing a soft-gated reasoning. Furthermore, to discriminate between semantically proximal emotions, we design a Contrastive Discriminative Visual Pruning (CDVP), which isolates discriminative visual tokens to reason the final emotion category by computing the absolute discrepancy between the attention distributions of similar emotions. Performances on several benchmarks demonstrate that DAN significantly improves discrimination for affective nuances without consuming additional training resources, especially achieving +10.47% improvements with Qwen3-VL-8B-Instruct on WebEmo25 dataset that contains 25 fine-grained emotion categories.