让多模态大模型根据输入动态调整回答，更安全准确。
Learning to Steer: Input-dependent Steering for Multimodal LLMs
- 用输入相关的线性偏移实现细粒度控制
- 减少幻觉，提升安全性和准确性
- 适合需要灵活响应的多模态应用
提示引导已成为在不修改模型的前提下调控大语言模型行为的有效方法。然而，该技术在多模态大语言模型（MLLMs）中仍研究不足；且现有方法（如均值引导）使用单一引导向量，与输入无关。这在行为依赖具体输入时面临局限：例如面对非法请求应拒绝回答，而医疗建议则需推荐外部资源或专家咨询。本文提出细粒度的输入相关引导，通过对比式输入特定提示计算线性偏移。由于测试时无法获取此类提示，我们引入一个小型辅助模块来预测输入相关的引导向量。所提方法L2S（Learn-to-Steer）在减少幻觉和强化安全性方面优于静态基线。代码已公开于 https://jayneelparekh.github.io/learn-to-steer/
原文摘要 · Abstract (English)
Steering has emerged as a practical approach to enable post-hoc guidance of LLMs towards enforcing a specific behavior. However, it remains largely underexplored for multimodal LLMs (MLLMs); furthermore, existing steering techniques, such as mean steering, rely on a single steering vector, applied independently of the input query. This paradigm faces limitations when the desired behavior is dependent on the example at hand. For example, a safe answer may consist in abstaining from answering when asked for an illegal activity, or may point to external resources or consultation with an expert when asked about medical advice. In this paper, we investigate a fine-grained steering that uses an input-specific linear shift. This shift is computed using contrastive input-specific prompting. However, the input-specific prompts required for this approach are not known at test time. Therefore, we propose to train a small auxiliary module to predict the input-specific steering vector. Our approach, dubbed as L2S (Learn-to-Steer), demonstrates that it reduces hallucinations and enforces safety in MLLMs, outperforming other static baselines. Our code is publicly available at https://jayneelparekh.github.io/learn-to-steer/
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。