video-supervision

Tag

Cards List
#video-supervision

SimVTA: Simulation-Grounded Video-Trajectory Alignment for Driving World Action Models

Hugging Face Daily Papers ↗ · 2026-10-09 Cached

SimVTA 是一个面向驾驶世界行动模型(WAM)的仿真落地后训练框架,通过 3D 高斯泼溅重建真实场景并渲染候选轨迹的视频监督,结合 GRPO 与流匹配损失优化规划能力,在 NAVSIM v1 上将 PDMS 从 86.2 提升至 91.3。

0 favorites 0 likes
← Back to home

Submit Feedback