Search papers, labs, and topics across Lattice.
2
0
5
3
Gemini Embedding 2's unified multimodal embeddings beat specialized models across diverse tasks and even generalize zero-shot to niche fields like astronomy and culinary arts.
Autonomous research agents are surprisingly unreliable, with existing systems hallucinating references 21% of the time and failing to align methods with code as often as 80%, but a new "Chain-of-Evidence" approach can fix this.
