1 paper
Mouxiang Chen, Hao Tian, Zhongxin Liu +2
While existing code large language models (code LLMs) exhibit impressive capabilities in code generation, their autoregressive sequential generation inherently lacks reversibility.…