引导对话式大语言模型开展长时情感支持对话
Steering Conversational Large Language Models for Long Emotional Support Conversations
AI总结：
本研究针对大语言模型长时情感支持对话中策略遵循性不足的问题，提出SRA评估指标与策略条件合成数据集，通过微调显著提升了Llama系列模型的策略可引导性。
摘要：
本研究旨在解决大语言模型（LLM）在长时对话中始终遵循情感支持策略的难题，聚焦Llama-2与Llama-3系列模型的可引导性，考察其在交互全程维持对应策略的能力。为评估该特性，我们提出策略相关注意力（SRA）指标，通过注意力图量化模型对提示策略的遵循程度。为支撑研究，我们基于ESConv数据集构建了一个以策略为条件的合成对话数据集。我们还基于所提SRA指标设计了多种基线方法以应对上述挑战，并提出一个微调模型，该模型显著提升了基础模型在对话全程遵循策略的可引导性。相关代码与数据已在我们的GitHub上公开。
英文摘要：
In this study, we address the challenge of enabling large language models (LLMs) to consistently adhere to emotional support strategies in extended conversations. We focus on the steerability of the Llama-2 and Llama-3 suite of models, examining their ability to maintain these strategies throughout interactions. To assess this, we introduce the Strategy Relevant Attention (SRA) metric, which quantifies the model's adherence to the prompted strategy through attention maps. To facilitate our study, we create a strategy-conditioned synthetic conversational dataset derived from the ESConv dataset. We also propose various baselines informed by our proposed SRA metric to address the challenge and propose a fine-tuned model that significantly enhances the steerability of the base model in following the strategy throughout the conversation. The code and data are publicly available on our GitHub.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
