Search papers, labs, and topics across Lattice.
Swinburne University of Technology
1
0
2
Agents that ace long-context recall can still bomb when they need to use that memory to actually *do* something, revealing a critical flaw in how we currently evaluate memory in AI.
