Benchmark Radar
AI BENCHMARK PROFILE

The Energy Society

General AIAgents

The Energy Society is a multi-agent simulation environment where LLM agents operate under energy constraints tied to token generation, completing jobs and donating energy to survive. The evaluation measures emergent cooperative and competitive behavior across different model sizes and incentive settings.

Released
2026-07-16
Readiness
Runnable
Primary field
General AI

Why it matters

The environment addresses the gap in studying how energy costs and survival pressures shape agent cooperation and competition, providing a testbed for evaluating multi-agent systems in resource-constrained settings. It could support decisions on agent design for sustainable and cooperative AI systems.

Motivation

LLM-based agents are increasingly deployed in multi-agent environments whose incentives can shape their behavior.

Primary resources

Benchmark Radar records only publicly supported details and links back to primary sources for verification.