VLM agents nail 94% of urban navigation but only 0.7% of hardest episodes end safely
lulzxdxdxd · reddit · 2026-10-14
An arXiv paper benchmarks eight vision-language-model agents on real-time urban navigation: they complete 94% of tasks overall, yet only 0.7% of the hardest episodes finish without a safety event — a stark gap between average competence and long-tail safety.
More from Research
- Revisiting STaR: models self-improve by training only on rationales that reach correct answers — AndrewLampinen · 2026-10-14
- Google researcher: having models write new training data via ICL improves finetuning generalization — AndrewLampinen · 2026-10-14
- Study: model collapse doesn't happen when synthetic data accumulates, even without new real data — AndrewLampinen · 2026-10-14
- Synthetic data cuts loss on real distribution and yields 1.8x data efficiency in constrained pretraining — AndrewLampinen · 2026-10-14
- AlphaNIST robot training: Newton physics plus sparse reward did the heavy lifting — milesmacklin · 2026-10-14
- Modal Trains RL on Mixed Accelerators, Cutting Cost While Matching Homogeneous Stability — charles_irl · 2026-10-14