Stair policy: 3 cm and 4 cm steps

Our own policy crosses the stair course in simulation, three steps up and three down on 80 mm treads. It was trained from our flat policy, and Stanford's policies served only as the baseline.

• 3 cm steps: 97 of 100 crossings on average, 91–100 over 20 evaluations (Stanford's climbing policy: 88, 83–93 over 15)
• 4 cm steps: 79 of 100 on average, 69–86 (Stanford's: 0–1)
• Standing height in the ROS 2 simulator: 0.176 m (official policy 0.183 m, our Oct 2 policy 0.136 m), no fall in 10 walk sequences
• Training: 1 B steps on flat ground, then 800 M steps on stairs of 2–4 cm
• Open: standing with a zero command, a front leg drifts to its joint limit in the training simulator, not in the ROS 2 simulator