Large Language Models for Program Synthesis

Dive into the future of programming with our upcoming talk on large language models (LLMs) and their impact on code generation. Join Xinyun Chen, a senior research scientist at Google DeepMind, as we navigate the promising yet challenging world of LLMs in tackling complex programming tasks and competitive programming problems.

📌 What We'll Cover:
The evolution of large-scale language models in code generation.
The achievements and limitations of current LLMs in understanding and solving complex programming tasks.
A detailed discussion on AlphaCode: how it ranks in the top 54.3% in Codeforces competitions, showcasing LLMs' potential in competitive programming.
Introducing Self-Debugging: a novel approach that enables LLMs to debug their own code through rubber duck debugging, significantly improving performance on code generation tasks.
How Self-Debugging optimizes sample efficiency and outperforms traditional models in text-to-SQL, code translation, and synthesizing Python functions from descriptions.

🎤 About Our Speaker:
Xinyun Chen brings a wealth of knowledge from the intersection of deep learning, programming languages, and security. With a Ph.D. from UC Berkeley, her pioneering research includes integrating SpreadsheetCoder into Google Sheets and featuring AlphaCode on the cover of Science Magazine.

🔗 Useful Links:

👉 Follow us on socials: