SafeLoop: Risk-Aware Rollback for Vision-Language-Action Manipulation
On the map: Vision-language-action models → Safety and adversarial attacks also fits: Failure recovery and robustness
Abstract (arXiv)
Recent vision-language-action (VLA) models are promising for general-purpose manipulation, but long-horizon execution remains fragile. Small state-estimation or control errors can lead to irreversible failures (e.g., collisions and object drops). Avoiding these risks requires a proactive safety mechanism capable of anticipating hazards. In this paper, we introduce SafeLoop, a non-invasive external wrapper that adds hazard prediction and rollback-based recovery to a VLA model without changing its parameters. SafeLoop trains a risk predictor from vision and proprioception to output four values: the probability and time-to-hazard for body collisions and for object failures. A lightweight controller then chooses one of three actions based on the predicted risk: continue execution (noop), save a safety checkpoint (record), or retreat in joint space (rollback). Rollback moves the robot back to a recent safe waypoint and queries the base policy again, which may yield an alternative continuation. Across 24 LIBERO tasks (16 random seeds each) and three real-robot tasks (25 rollouts each), SafeLoop achieves a stronger overall safety-success trade-off than alternative methods, reducing hazard cases by roughly 70% while preserving task success and the base-policy control rate. Project code is available at https://github.com/Loule0-0/SafeLoop/tree/release/safeloop.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Safety and adversarial attacks, one of 13 directions in Vision-language-action models. Builds on 15 of its 43 references that are in the map. First posted on arXiv 22 Sep 2026.
Builds on
15 of 43 references.
Closest ancestors · 1
Same direction, newest first
Other references · 14
- 2017Proximal Policy Optimization Algorithmscited by 14,441
- 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
- 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
- 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,007
- 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
- 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Successcited by 844
- 2025WorldVLA: Towards Autoregressive Action World Modelcited by 351
- 2025π*0.6: a VLA That Learns From Experiencecited by 280
- 2018Safe Exploration in Continuous Action Spacescited by 237
- 2025A Careful Examination of Large Behavior Models for Multitask Dexterous Manipulationcited by 125
- 2025RynnVLA-002: A Unified Vision-Language-Action and World Modelcited by 70
- 2025Agentic Robot: A Brain-Inspired Framework for Vision-Language-Action Models in Embodied Agentscited by 30
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026VLPSA: Vision-Language-Poisson-Safe Actions for Full-Body Safety of Learned Policies
- 2026SafeVLA-Bench: A Benchmark for the Success-Safety Gap in Vision-Language-Action Models
- 2026SafeStage: Evaluating Safety Before, During, and After Vision-Language-Conditioned Robot Manipulation ↗
- 2026LIBERO-Safety: A Comprehensive Benchmark for Physical and Semantic Safety in Vision-Language-Action Models
- 2026ForesightSafety-VLA: A Unified Diagnostic Safety Benchmark for Vision-Language-Action Models
- 2025SafeVLA: Towards Safety Alignment of Vision-Language-Action Model via Constrained Learning
- 2025SAFE: Multitask Failure Detection for Vision-Language-Action Models
- 2026ReconVLA: An Uncertainty-Guided and Failure-Aware Vision-Language-Action Framework for Robotic Control