Akxar Labs
About
An early-stage research lab studying how AI agents behave in open-ended, multi-agent settings.
Standard evaluations tend to examine short trajectories of a single model. Open-ended environments with many agents can reveal more: long-horizon coordination, mutual influence, and strategies no one anticipated.
Akxar Labs designs scenarios that probe specific behavioral dispositions, builds monitoring and oversight tooling for multi-agent systems, and aims to make the underlying environments available as a shared resource for safety research.
Get in touch
We are at an early stage and welcome conversations with researchers and collaborators. Reach us at sneheel@akxar.xyz.