WOD end-to-end challenge

Qwen released Qwen-Drive-1.0-4B, a 4-billion-parameter vision-language model built on Qwen3.5-4B that fuses perception and planning into a single model with a bird's-eye-view perception head and two p…