When Visual Evidence Is Insufficient: How Instructions, Post-Training, and Internal Activations Shape Refusal in Multimodal Language Models
作者: Ananth K Suresh, Aveen Dayal, Gouthaman KV, Aditya Pola, Vineeth N Balasubramanian
分类: cs.LG, cs.CV
发布日期: 2026-10-27
备注: 31 pages, 6 figures
💡 一句话要点
提出多模态语言模型拒绝机制的优化策略
🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)
关键词: 多模态语言模型 拒绝机制 推理时指令 后训练 内部激活干预 视觉证据识别 模型优化
📋 核心要点
- 现有多模态语言模型在识别图像证据不足时的拒绝能力不足，可能导致不准确的回答。
- 论文提出通过推理时指令、后训练和内部激活干预来优化模型的拒绝机制，以提高其对视觉证据的敏感性。
- 实验结果表明，拒绝行为对指令的意义和位置敏感，后训练显著提高拒绝率，但也可能导致混合回答。
📝 摘要（中文）
为了可靠地部署多模态大型语言模型（MLLMs），模型需识别图像证据不足的情况并拒绝不支持的回答。然而，拒绝并不一定表明模型识别了缺失的证据。本文研究了推理时指令、后训练和内部激活如何影响模型在视觉上无法回答的问题上的拒绝行为。通过在MoHoBench基准上对三种MLLMs进行系统性实验，发现拒绝行为对指令的意义和位置高度敏感，且后训练可以显著提高拒绝率，但也可能导致拒绝语言与尝试回答的混合。研究结果强调了外部指令、后训练和内部表示共同塑造拒绝行为的重要性。
🔬 方法详解
问题定义：本文解决多模态语言模型在视觉证据不足时的拒绝能力不足的问题。现有方法可能因安全约束或未识别缺失证据而拒绝回答，导致不可靠的输出。
核心思路：通过系统性地操控推理时指令的意义和位置、比较监督与偏好基础的后训练，以及干预与拒绝相关的内部激活，来优化模型的拒绝机制。
技术框架：整体架构包括三个主要模块：推理时指令的设计与实施、后训练策略的比较、以及内部激活的干预。每个模块针对不同的拒绝机制进行优化。
关键创新：最重要的创新在于识别并区分外部指令、后训练和内部表示对拒绝行为的影响，强调了拒绝行为的复杂性与多样性。
关键设计：在推理时指令中，采用了证据感知与一般拒绝指令的对比，后训练中则关注于拒绝语言与尝试回答的结合，内部激活干预则确保在视觉可回答样本上不出现无差别拒绝。
🖼️ 关键图片
📊 实验亮点
实验结果显示，证据感知和一般拒绝指令的效果显著优于不匹配的安全指令。后训练显著提高了拒绝率，但也引入了拒绝语言与尝试回答的混合。内部激活干预则在不同提示条件下产生了可转移的变化，未在视觉可回答样本上表现出无差别拒绝。
🎯 应用场景
该研究的潜在应用领域包括智能助手、自动问答系统和图像理解等。通过优化多模态语言模型的拒绝机制，可以提高其在实际应用中的可靠性和用户体验，未来可能推动更智能的交互系统的发展。
📄 摘要（原文）
For reliable deployment, multimodal large language models (MLLMs) should recognize when an image provides insufficient evidence to answer a question and withhold unsupported answers. Yet refusal alone is not necessarily evidence of such recognition: models may refuse because of unrelated safety constraints or without identifying the missing evidence. We study how refusal on visually unanswerable questions is shaped by three sources of control: inference-time instructions, post-training, and internal activation interventions. Across three MLLMs on the MoHoBench benchmark, we systematically manipulate the meaning and placement of refusal instructions, compare supervised and preference-based post-training, and intervene on a refusal-related activation direction derived from naturally answered and refused examples. We find that refusal is highly sensitive to both the meaning and placement of instructions, with model-specific changes at the sample level that are obscured by aggregate refusal rates. Evidence-aware and general refusal instructions have stronger effects than an intentionally mismatched safety instruction. Post-training produces larger increases in refusal, but can also induce responses that combine refusal language with attempted answers. In contrast, activation intervention produces smaller yet transferable changes across prompt conditions, without evidence of indiscriminate refusal on tested visually answerable samples. These findings show that refusal is jointly shaped by external instructions, post-training, and internal representations, highlighting the need to disentangle evidence recognition from a model's learned propensity to refuse.