Collaborative Reasoner

Collaborative Reasoner - Self-Improving Social Agents with Synthetic Conversations

Collaborative Reasoner (Coral), a framework for evaluating and enhancing the collaborative reasoning skills of Language Models (LLMs) in multi-turn conversational settings, which are often overlooked in single-turn evaluations. Coral addresses the challenge of limited conversational data by proposing a self-improvement method that utilizes Matrix, a scalable multi-agent communication framework, to generate synthetic interaction data through simulated self-collaborations of LLMs.