Beyond Partner Diversity: An Influence-Based Team Steering Framework for Zero-Shot Human-Machine Teaming
The paper presents a new framework called Influence-Based Team Steering (IBTS) for enhancing human-machine teaming in zero-shot coordination scenarios. It addresses the limitations of existing data-driven methods by focusing on influence shaping to improve team interaction patterns. The evaluation of IBTS shows improved performance in team settings, emphasizing the importance of combining coordination mechanisms with partner variation.
- ▪The framework IBTS aims to enhance human-machine teaming by using influence shaping.
- ▪Zero-shot coordination (ZSC) is proposed to simulate diverse partner behaviors without extensive human interaction data.
- ▪IBTS was tested in Overcooked-AI with both simulated and real human teammates, showing improved team performance.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.15400 (cs) [Submitted on 14 May 2026] Title:Beyond Partner Diversity: An Influence-Based Team Steering Framework for Zero-Shot Human-Machine Teaming Authors:Wei Sheng, Rohan Paleja View a PDF of the paper titled Beyond Partner Diversity: An Influence-Based Team Steering Framework for Zero-Shot Human-Machine Teaming, by Wei Sheng and 1 other authors View PDF HTML (experimental) Abstract:While AI agents are rapidly advancing from isolated tools to interactive collaborators, data-driven human-machine teaming (HMT) methods remain costly in their reliance on human interaction data across domains, teammates, and team sizes.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
