Open-World Neural Navigation
Real-time neural maps for unfamiliar indoor environments.
Highlights
Compositional Vision-Language Reasoning
A training objective that improves compositional generalization.
Long-Form Video Memory Transformers
Memory-augmented transformers reasoning over hour-long videos.
Self-Supervised Medical Imaging
Cutting radiology annotation cost by 10x via contrastive pretraining.
Diffusion Policies for Manipulation
Diffusion action models for bimanual robot hands.
Causal Discovery from Video
Learning causal graphs directly from raw video.
On-Device Multimodal Inference
Running 7B-parameter VLMs on a single mobile GPU.