RACE: A Rollout-Free Metric for Selecting Robot Policy Checkpoints
What happened
Changyeon Kim and co-authors propose RACE, a ranking metric for robot policy checkpoints that requires no real-robot rollout: using a fixed set of held-out success trajectories, it performs state-conditioned threshold calibration on action deviations, then uses multiplicative accumulation to downweight the influence of large deviations on the reference prediction. The authors report that in simulation RACE outperforms baselines across different validation sources, with some baselines degrading markedly when the validation set changes; on real robots, its ranking is consistent with deployment success rates across the evaluated policy backbones and tasks of varying difficulty.
AI 根据报道生成 · 6 days ago更新
Reporting timeline
Follow the reporting to explore different perspectives.
- arXiv cs.RORACE: Rollout-Free Calibrated Cumulative Scoring for Robot Policy Checkpoint Selection
The study proposes RACE, a robot policy checkpoint scoring metric that requires no real-world rollout. It uses a fixed set of held-out successful trajectories to calibrate action deviations against state-conditional thresholds, then applies a running product to reduce the weight of later reference predictions after large deviations. In simulation, RACE outperformed baselines across different validation sources, while some baselines degraded noticeably when the validation set changed; on a real robot, its rankings were consistent with deployment success rates across the evaluated policy backbones and tasks of varying difficulty.
Activity for this event
- Current
- 3
- Peak
- 10Nov 5 14:00
- Past 24 hours
- -49%
The trend compares participants observed throughout the same period, so its range may be smaller than the current interest count. Move or tap to inspect each hour; use the arrow keys on a keyboard.