Steering Conversational Large Language Models for Long Emotional Support Conversations
This work addresses the problem of maintaining emotional support consistency in AI conversations for mental health applications, though it is incremental as it builds on existing models and datasets.
The study tackled the challenge of making large language models consistently follow emotional support strategies in long conversations by introducing the Strategy Relevant Attention metric and a fine-tuned model, which significantly improved steerability over base models like Llama-2 and Llama-3.
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.