Level 02
Dodging (proof of concept)
Train the bot live: reward useful actions, punish bad ones, and survive falling hazards.
AI acts once per second. Controls: D reward · A punish · R reset
Status: running
Time: 0.0s
Turn: 0
Score: 0
Best: 0
Last action: idle
Feedback token: spent
Action weights
- left: 1.00
- idle: 1.00
- right: 1.00
death
escape