We are creating a reinforcement learning projects.
Multi-agent RL sim: 4 traders, market maker & SEC regulator trained via GRPO on Llama-3.2. 250 steps of emergent financial crisis — slaughter, adaptation, collusion, regulatory oversight — no scripts, just learned behavior.
.png&w=256&q=75)
.png&w=640&q=75)