Search papers, labs, and topics across Lattice.
3
0
5
35
Training agents in deep, evolving environments can dramatically enhance their performance, with a 9B model achieving a 30.6% accuracy increase through targeted design.
Reducing visual token usage by 46% while improving performance shows that CUAs can leverage more historical data effectively without overwhelming compute budgets.
Process-level evaluation reveals hidden skill discrepancies among web agents, enabling targeted improvements that traditional success metrics overlook.
