Small Language Models vs. LLMs for Multi-Turn Customer-Service QA

A new study evaluates instruction-tuned Small Language Models (SLMs) for multi-turn customer-service question answering, comparing them against Large Language Models (LLMs). The research uses a history summarization strategy to preserve conversational context and introduces a conversation stage-based qualitative analysis. Nine low-parameterized SLMs are tested against three commercial LLMs. The findings aim to determine whether SLMs can effectively handle context-summarized multi-turn QA in resource-constrained environments, offering a more efficient alternative to computationally expensive LLMs.