Track new results across pure and applied mathematics.
Search papers, labs, and topics across Lattice.
2
0
6
Even the most advanced multi-task policies struggle to achieve 30% success on a new benchmark designed to test generalizable dexterous manipulation, revealing critical gaps in current robotic intelligence.
Frontis-MA1 achieves a remarkable 71.21% Medal Average on MLE-Bench Lite, outperforming leading models and showcasing the potential of AI systems to recursively improve their own engineering processes.