Search papers, labs, and topics across Lattice.
2
0
5
1
Even GPT-5.1 struggles to distinguish AI-generated academic images from real ones, achieving only 48.8% accuracy, revealing a significant gap between generative and forensic AI capabilities.
Traditional research papers are costing AI agents reproducibility and understanding, but a new "Agent-Native" format that captures the full messy research process boosts performance by up to 20%.
