From Task Solvers to Teammates: A Theory-Grounded Architecture for Advancing Collaboration Readiness in LLM Agents

This project reimagines AI agents not just as autonomous problem-solvers but as effective collaborators. It introduces a theory-grounded approach to design and evaluate Large Language Model agents for human–AI teamwork. The research develops CollabBench, the first benchmark to measure collaboration readiness using metrics beyond task success, such as communication efficiency and coordination quality. It also proposes a modular Collaborative Readiness Layer architecture that externalizes key functions like common ground and workspace awareness, making agent behaviour more transparent and controllable. Together, these innovations aim to transform AI agents into trusted digital teammates, enabling safer and more productive human–agent collaboration.

This research is conducted via The Agentic AI Research and Innovation (AARI) Initiative which focuses on the next frontier of agentic systems through Grand Challenges with the academic community and Microsoft Research.