Search papers, labs, and topics across Lattice.
3
0
5
1
LLMs can debug code more effectively by reasoning about intermediate program states, not just relying on coarse pass/fail test signals.
LLMs can fix more bugs when guided by a semantic evolutionary algorithm that recombines successful partial fixes across a population of candidate repairs.
By explicitly modeling code structure and dependencies, Hydra lets smaller code LLMs beat much larger models on repository-level code generation tasks.
