ChatGPT Pro Improves Capacity Bounds for Binary Deletion Channel by 4%

DimitrisPapail · x · 2026-08-02

A researcher tested ChatGPT Pro on a classic open problem in information theory: the capacity bounds of the binary deletion channel. At a 50% deletion probability, the model successfully derived a new result, improving the best current lower bound by about 4%. This demonstrates the potential of LLMs in assisting advanced pure mathematics research.

Related event: ChatGPT Solves 15-Year Information Theory Puzzle(2 posts)→

Original post →

More from Research

Research channel →