highway state rendererGPT-5.5
Codex- Held-out 分数
- -30.24
- Checkpoint
- #012
- 重跑
- 220 steps
Validation-selected checkpoint rerun in the original environment; case return -48.79.
Driving · Core16
把车辆连续控制到目标停车姿态。
这些证据属于论文时期实验,并不表示相关 package 当前仍然可用。
每张卡片都在原研究 Environment 中重跑 validation 选出的 checkpoint。
highway state rendererValidation-selected checkpoint rerun in the original environment; case return -48.79.
highway state rendererValidation-selected checkpoint rerun in the original environment; case return -46.70.
highway state rendererValidation-selected checkpoint rerun in the original environment; case return -45.59.
highway state rendererValidation-selected checkpoint rerun in the original environment; case return -24.69.
在该 Environment 内分数越高越好;不同任务的原始 reward scale 不可比较。
-30.24-32.71-38.18-47.45-53.38