Search papers, labs, and topics across Lattice.
2
0
3
4
Despite advances in agentic systems, they still lag far behind human performance in complex tasks, achieving only 19.1% success on a new benchmark designed to probe their limitations.
Current multimodal agents navigating document collections achieve human-level accuracy through brute-force search, highlighting a critical gap in strategic reasoning and planning.
