Search papers, labs, and topics across Lattice.
1
0
2
Tool-using LLMs face a near-universal robustness gap, but combining Bayesian Tool Memory with reinforcement learning can boost recovery performance by over 40% in failure scenarios.
