Manmohan Chandraker

LDR achieves unprecedented extrapolation of learned dynamics in video models, outperforming traditional methods by a staggering margin in both accuracy and efficiency.

Task-aligned simulated futures can dramatically improve robot policy learning, yielding superior performance in complex manipulation tasks.

SliceScorer reveals critical coverage gaps in driving VLMs that traditional methods overlook, enhancing safety validation in real-world applications.

Control over physical properties like friction and restitution in generated videos is now possible, paving the way for more realistic and controllable video synthesis.

Editing driving scenes with language just got a whole lot better: HorizonWeaver lets you scalably generate photorealistic, controllable scenarios, outperforming existing methods by a wide margin.

Real-time autonomous driving with language models is now possible, achieving 3x speedup and state-of-the-art performance by combining learned and rule-based planning.