Enregistré dans:
| Auteurs principaux: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | https://arxiv.org/abs/2603.10282 |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866918534824591360 |
|---|---|
| author | Attarian, Maria Vyse, Ian Voelcker, Claas Gerigk, Jasper Opryshko, Evgenii Almasri, Anas Singh, Sumeet Du, Yilun Gilitschenski, Igor |
| author_facet | Attarian, Maria Vyse, Ian Voelcker, Claas Gerigk, Jasper Opryshko, Evgenii Almasri, Anas Singh, Sumeet Du, Yilun Gilitschenski, Igor |
| contents | In recent years, Behavior Cloning (BC) has become one of the most prevalent methods for learning manipulation from human demonstrations. Despite their successes, BC policies are often brittle and struggle with precise manipulation. To overcome these issues, we propose UF-OPS, an Update-Free On-Policy Steering method that enables the robot to predict the success likelihood of its actions and adapt its strategy at execution time. We accomplish this by training verifier functions using policy rollout data obtained during an initial evaluation of the policy. These verifiers are subsequently used to steer the base policy toward actions with a higher likelihood of success. Our method improves the performance of black-box diffusion policies, without changing the base parameters, making it lightweight and flexible. We present results from both simulation and real-world data and achieve an average 49% improvement in success rate over the base policy across 5 real tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_10282 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Update-Free On-Policy Steering via Verifiers Attarian, Maria Vyse, Ian Voelcker, Claas Gerigk, Jasper Opryshko, Evgenii Almasri, Anas Singh, Sumeet Du, Yilun Gilitschenski, Igor Robotics In recent years, Behavior Cloning (BC) has become one of the most prevalent methods for learning manipulation from human demonstrations. Despite their successes, BC policies are often brittle and struggle with precise manipulation. To overcome these issues, we propose UF-OPS, an Update-Free On-Policy Steering method that enables the robot to predict the success likelihood of its actions and adapt its strategy at execution time. We accomplish this by training verifier functions using policy rollout data obtained during an initial evaluation of the policy. These verifiers are subsequently used to steer the base policy toward actions with a higher likelihood of success. Our method improves the performance of black-box diffusion policies, without changing the base parameters, making it lightweight and flexible. We present results from both simulation and real-world data and achieve an average 49% improvement in success rate over the base policy across 5 real tasks. |
| title | Update-Free On-Policy Steering via Verifiers |
| topic | Robotics |
| url | https://arxiv.org/abs/2603.10282 |