Search papers, labs, and topics across Lattice.
2
0
3
0
Language models struggle to recover research ideas from bibliographies, achieving only 3-15% accuracy, but a multi-agent approach boosts this to 23-42%.
Current AI research agents miss critical metacognitive checks, leading to pervasive failures across diverse scientific tasks.
