FAR: Failure-Aware Retry for Test-Time Recovery and Continual Policy Improvement
Overview
FAR enables robots to learn from failures, adapt behavior, and complete tasks autonomously. Experiments show improved success rates and data efficiency in simulation and real-world settings. Videos and code available at the provided link. arXiv:2607.01111v2 Robot policies inevitably encounter failures when deployed in real environments. FAR combines Failure-Contrastive Preference Adaptation with lightweight action perturbations to steer policies away from unsuccessful behaviors and encourage local exploration. FAR improves data efficiency during continual policy improvement by exploiting informative failure cases. arXiv:2607.01111v2 Experiments in both simulation and real-world manipulation tasks show that FAR substantially improves success rates and robustness. Videos and code are available at https://hoar012.github.io/FAR-Project. arXiv:2607.01111v2
Generated from attributed reports · Updated 2 hours ago
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
No current evidence-backed claims. Missing information remains not reported.
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsFAR: Failure-Aware Retry for Test-Time Recovery and Continual Policy Improvement
This paper proposes FAR, a framework that allows robots to learn from failures during deployment, adapt their behavior, and complete tasks autonomously. FAR combines failure-contrastive preference adaptation with lightweight action perturbations to encourage local exploration and integrates successful recovery trajectories into a training loop for continual policy improvement. Experiments show substantial gains in success rates and data efficiency in both simulation and real-world settings.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).