Nitish
fix: clamp ALL score outputs to (0.01, 0.99) β€” inference.py score + environment total_reward
c04a5c5