RT-SAFE benchmark: frontier VLMs hit 94% task success but only 0.7% safe completion

SimWorld's RT-SAFE benchmark evaluates embodied agents in continuously changing environments. Frontier VLMs achieve up to 94% task success, but real-time safety success drops to just 0.7%, exposing a huge safety gap.

2026-10-06 ~ 2026-10-06 · 3 related posts

• RT-SAFE benchmark: 94.1% of 8 frontier VLMs reach goals, only 0.7% finish with zero safety events — Lianhuiq · 2026-10-06
• RT-SAFE benchmark: frontier VLMs hit 94.1% task success but only 0.7% finish safely — Lianhuiq · 2026-10-06

1 near-duplicate retellings: Lianhuiq