Accepted Papers

Oral Presentations

1. Perturbation-Based Uncertainty for Failure Detection in Vision-Language-Action Models
2. Paired Action-Intervention Audits for Planner-Facing Embodied Model Safety
3. Decomposing VAM Uncertainty: Training-Free and Supervised Probes for Robot Failure Detection
4. Red-Teaming of Vision-Language-Action Models via Quality Diversity Environment Generation
5. Safer Imitation by Construction: Cutting Robot Failures at the Behavioral Fork

Posters

1. The Recovery Margin Is a Safety Margin: Auditing What Saturated Benchmarks Hide in Vision-Language-Action Policies
2. Falsifying Simulator Safety Claims with Action-Cost Baselines
3. Dormant but Not Latent in Robot Policies
4. Route Guided CBF-QP Repair for Safe Vision Language Action Models
5. Future Directions for Quality Diversity Red-Teaming of Vision-Language-Action Models
6. SafeCasa: A Benchmark Dataset for Physical Safety in Household Robotic Manipulation
7. Data-Efficient Verification of Black-box Controllers via Flow Matching and Conformal Inference
8. Discriminative Barrier Functions for Safe Adversarial Imitation Learning from Observation
9. Defining and Auditing Dependency-Inducing Human-Robot Interaction
10. Safety Starts in the Dataset: Auditing Temporal Action-Label Misalignment Before Training
11. You Cannot Monitor Your Way Out of a Design Deficit: Inspectable VLA Internals as a Shared Safety Substrate