Question 1: Short answer
hardAI-checkedMulti-Agent SystemsIn a cooperative task all agents receive one shared team reward, and COMA gives each agent its own advantage through a counterfactual baseline. Agent has three actions. With the other agents' actions held fixed, the centralised critic and agent 's policy give:
| agent 's action | |||
|---|---|---|---|
| 4.0 | 6.0 | 1.0 | |
| 0.2 | 0.5 | 0.3 |
Agent actually took . What is its COMA advantage?