Kunhao Zheng

Execution time can be transformed into a learnable reward, leading to substantial improvements in code optimization performance in RL settings.

Extrapolating between code-generating RL agents trained on different unit test coverages unlocks better correctness-efficiency trade-offs than any single agent alone.

LLMs can now automatically verify imperative code at scale, achieving state-of-the-art results on challenging verification benchmarks and paving the way for large-scale verified code datasets.