Failure Prediction and Classification in Vision-Language-Action Models Via Calibrated Latent Probing
cs.RO, cs.AI
Submitted: 2026-11-05
Updated: 2026-11-05
Terminology
Sources
- Unpacking Failure Modes of Generative Policies: Runtime Monitoring of Consistency and Progress
- Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Control
- Know Where You're Uncertain When Planning with Multimodal Foundation Models: A Formal Framework
- Observing and Controlling Features in Vision-Language-Action Models
- AEGIS: A Backup Reflex for Physical AI
- SentryVLA: Uncertainty-Guided Action Selection and State Monitoring for Vision-Language-Action Robotic Control
- DREAMSTEER: Latent World Models Can Steer VLA Policies During Deployment Without Any Finetuning
- LIBERO-Plus: In-depth Robustness Analysis of Vision-Language-Action Models
- Temporal Difference Calibration in Sequential Tasks: Application to Vision-Language-Action Models
- SAFE: Multitask Failure Detection for Vision-Language-Action Models
- ActProbe: Action-Space Probe for Early Failure Detection of Generative Robot Policies
- Learning What to Say to Your VLA: Mostly Harmless Vision Language Action Model Steering
- VLAs are Confined yet Capable of Generalizing to Novel Instructions
- LIBERO-Occ: Evaluating and Improving Vision-Language-Action Models under Scene-Induced Occlusion via Viewpoint Imagination
- LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learning
- VLS: Steering Pretrained Robot Policies via Vision-Language Models
- CheckVLA: Execution-Time Verification with Action-Conditioned World Model for Long-Horizon Mobile Manipulation
- GR00T N1: An Open Foundation Model for Generalist Humanoid Robots
- Hide-and-Seek in Trajectories: Discovering Failure Signals for VLA Runtime Monitoring
Related papers
- Rethinking Actions: Continuity-Aware Priors for World-Model Planning
- MOSAIC: Modular Obstacle and Spatial-Aware Framework for Manipulation In Cluttered Scenes
- SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
- DexSeed: Expanding Human Hand-Object Interactions into Robot Demonstrations for Dexterous Policy Learning
- LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning
- CIALLO: Learning Shared Intent Representations for Cross-Embodiment Manipulation