Reflect-VLA: Internalizing Reflective Guidance via Dual-Path On-Policy Self-Distillation for Autonomous Driving

Benchmark Model Rank Results
bench2drive-on-bench2driveReflect-VLA-0.8B–Driving Score: 76.03
bench2drive-on-bench2driveReflect-VLA-2B–Driving Score: 77.39
bench2drive-on-bench2driveReflect-VLA-4B–Driving Score: 80.15
navsim-on-opensceneReflect-VLA-0.8B–PDMS: 89.3
navsim-on-opensceneReflect-VLA-2B–PDMS: 90.0
navsim-on-opensceneReflect-VLA-4B–PDMS: 91.5