Search papers, labs, and topics across Lattice.
1
0
3
2
Most multimodal models fail to truly understand tool use in real-world egocentric videos, with even Gemini-3.1-Pro struggling to ground actions in visual evidence.
