← Back to Benchmarks
simmediumrlmetric · varies

Training Generalizable Collaborative Agents via Strategic Risk Aversion

Description

Many emerging agentic paradigms require agents to collaborate with one another (or people) to achieve shared goals. Unfortunately, existing approaches to learning policies for such collaborative problems produce brittle solutions that fail when paired with new partners. We attribute these failures to a combination of free-riding during training and a lack of strategic robustness. To address these problems, we study the concept of strategic risk aversion and interpret it as a principled inductive

Source

http://arxiv.org/abs/2602.21515v2