Search papers, labs, and topics across Lattice.
3
0
5
Tactile-reactive manipulation can boost robotic success rates by over 30% in delicate tasks, revolutionizing how robots interact with their environment.
Stateful Visual Encoders enable VLMs to leverage prior visual context, leading to substantial performance gains in multi-image tasks.
Despite advancements in audio-language models, their ability to reliably "hear" pitch is surprisingly poor, raising concerns about their suitability for music-related tasks and multimodal AI applications.
