Competition-level code generation with AlphaCode

ABSTRACT

Programming represents a complex problem-solving task that requires deep logic and algorithmic reasoning. We present AlphaCode, a system that writes computer programs at a competitive level. AlphaCode uses transformer-based language models to generate millions of diverse code samples for a given problem, which are then filtered and clustered to select a small set of submissions. When evaluated on the Codeforces platform against human competitors in unseen programming challenges, AlphaCode achieved an estimated rank within the top 54% of participants. This marks the first time an AI system has reached a competitive level in programming competitions, highlighting a significant leap in machine reasoning.