Search papers, labs, and topics across Lattice.
2
0
4
5
AgentEval uncovers up to 38 hidden failure boundaries in conversational LLMs that traditional testing methods overlook.
LLMs can now generate significantly better Java unit tests without mocks, achieving up to 25% higher branch coverage by learning from existing code and rigorously enforcing semantic constraints.
