Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
In this invited talk from POPL 2025, Professor Loris D'Antoni from the University of California, San Diego explores the evolution of program synthesis over the past 15 years and introduces innovative approaches to finding good programs by eliminating provably incorrect ones. Discover the contrast between traditional symbolic/enumerative methods and modern large language models (LLMs) in code generation, focusing on their different specifications, guarantees, and approaches. Learn about techniques for proving unrealizability—determining when a set of programs contains only incorrect solutions that synthesizers should discard. Explore recent research on constraining LLM outputs to follow user intents and understand how unrealizability serves as a bridge between formal methods and AI-based code generation. This hour-long presentation highlights exciting opportunities for programming language researchers to influence the future of AI-assisted programming, delivered as part of the ACM SIGPLAN POPL 2025 conference.
