SafeDriveVLA (CoRL 2026): world-model dreaming catches unsafe driving instructions before acting
rsasaki0109 · x · 2026-10-13
A CoRL 2026 paper tackles a blind spot in driving VLAs: what if a free-form navigation instruction is unsafe? SafeDriveVLA introduces benchmarks covering instruction following, scene conflicts, and adversarial robustness, and combines explicit driving modes with world-model "dreaming" (InternVL3-1B + V-JEPA 2 encoder) to evaluate maneuvers before executing them.
Related event: SafeDriveVLA Uses World-Model Rehearsal to Block Unsafe Driving Commands(2 posts)→
More from Embodied
- Perseverance rover breaks off-world driving record at 28.10 miles — NASA · 2026-10-13
- ASCII Depth-Perception Art With a Mac and RealSense D435 Camera — chrismatthieu · 2026-10-13
- SFGate slams Zuckerberg's new Muse gadget: 'a parasite, don't fall for it' — sfgate · 2026-10-13
- Leaked NVIDIA slides reveal Factory AI blueprint with Nemoclaw agent framework — firstadopter · 2026-10-13
- Berkeley's QF3 makes RL work for flow policies on real robots with one clipped gradient tweak — andrew_n_carr · 2026-10-13
- Google DeepMind Robotics is hiring a pretraining & data quality research scientist in London — bousmalis · 2026-10-13