List your company

Strategic Interaction Optimization (Constant Response) on Indirect Reciprocity Degenerate Regimes

95Reference Performancereciprocity gradient
1737.2557.577.75May 8, 2026

Evaluation results

MethodLinks
reciprocity gradientRepresentative opponents=ProudCoop+AllD (signal, flat-coop attractor), Required best response=constant cooperate at b=102026.0595
Model-free baselines (DPG/DDPG/TD3)Representative opponents=L3 opponents, Training Domain=signal, Required best response=constant φ⁰ matches the norm2026.0592
reciprocity gradientRepresentative opponents=Identity opponents, Training Domain=action, Required best response=constant π⁰ = 12026.0589
reciprocity gradientRepresentative opponents=L3 opponents, Training Domain=action, Required best response=constant π⁰ = 12026.0589
reciprocity gradientRepresentative opponents=L3 opponents, Training Domain=signal, Required best response=constant φ⁰ matches the norm2026.0584
Model-free baselines (DPG/DDPG/TD3)Representative opponents=L6 opponents, Training Domain=signal, Required best response=constant φ⁰ matches the norm2026.0579
Model-free baselines (DPG/DDPG/TD3)Representative opponents=ProudCoop+AllD (signal, flat-coop attractor), Required best response=constant cooperate at b=102026.0579
reciprocity gradientRepresentative opponents=L6 opponents, Training Domain=signal, Required best response=constant φ⁰ matches the norm2026.0571
reciprocity gradientRepresentative opponents=Identity opponents, Training Domain=signal, Required best response=constant φ⁰→12026.0567
Model-free baselines (DPG/DDPG/TD3)Representative opponents=Identity opponents, Training Domain=signal, Required best response=constant φ⁰→12026.0553
Model-free baselines (DPG/DDPG/TD3)Representative opponents=L3 opponents, Training Domain=action, Required best response=constant π⁰ = 12026.0533
Model-free baselines (DPG/DDPG/TD3)Representative opponents=Identity opponents, Training Domain=action, Required best response=constant π⁰ = 12026.0520