SOTOPIA-$Ω$: Dynamic Strategy Injection Learning and Social Instruction Following Evaluation for Social Agents
This work addresses the gap in integrating human social strategies into social agents, enhancing their capabilities for applications like negotiation and dialogue, though it appears incremental as it builds on existing frameworks and strategies.
The paper tackled the problem of transferring human social strategies to language agents by proposing the SOTOPIA-Ω framework, which dynamically injects multi-step reasoning and direct strategies to automate high-quality social dialogue corpus construction, and demonstrated that trained 7B models significantly surpass GPT-4 in achieving social goals and improve Social Instruction Following performance.
Despite the abundance of prior social strategies possessed by humans, there remains a paucity of research dedicated to their transfer and integration into social agents. Our proposed SOTOPIA-$Ω$ framework aims to address and bridge this gap, with a particular focus on enhancing the social capabilities of language agents. This framework dynamically injects multi-step reasoning strategies inspired by negotiation theory and two simple direct strategies into expert agents, thereby automating the construction of a high-quality social dialogue training corpus. Additionally, we introduce the concept of Social Instruction Following (S-IF) and propose two new S-IF evaluation metrics that complement social capability. We demonstrate that several 7B models trained on high-quality corpus not only significantly surpass the expert agent (GPT-4) in achieving social goals but also enhance S-IF performance. Analysis and variant experiments validate the advantages of dynamic construction, which can especially break the agent's prolonged deadlock.