Search papers, labs, and topics across Lattice.
Address:
4
0
6
Static knowledge isn't enough; LLM agents achieve only a 72.6% success rate in executing complex aviation tasks, revealing significant gaps in procedural execution.
Current robot manipulation benchmarks fail to capture the messy reality of real-world deployment, so this work introduces a new benchmark, ManipArena, to close the sim2real gap.
Unlock the power of web videos for embodied AI: implicit geometry representations let agents learn to navigate from real-world room tours without relying on fragile 3D reconstruction.
Visuomotor policies can learn to ignore distracting visual variations simply by preprocessing raw RGB images into task-aware, semantic-geometric representations *before* feeding them to the policy.
