BenchmarkAlpasim 仿真器
Alpasim
An autonomous-driving simulator for closed-loop evaluation of VLA models and driving policies, supporting both open-loop and closed-loop testing.
Timeline (last 90 days)
Oct 30
In the Alpasim driving simulator, Q-adaptive guidance outperforms all fixed-weight schemes as well as online reinforcement learning methods (SAC, GRPO) in the open-loop setting.
Unverified50%
Oct 30
In Alpasim's closed-loop testing, Q-adaptive guidance passed 262 of 300 scenarios, while the original policy passed 252.
Unverified50%
All Facts (2)
Unverified
In the Alpasim driving simulator, Q-adaptive guidance outperforms all fixed-weight schemes as well as online reinforcement learning methods (SAC, GRPO) in the open-loop setting.
50%UnverifiedIn Alpasim's closed-loop testing, Q-adaptive guidance passed 262 of 300 scenarios, while the original policy passed 252.
50%