ChatGPT Cracks 30-Year-Old Graph Theory Problem
Key Takeaways
- ▸ChatGPT discovered a counter-example disproving the Dinitz–Garg–Goemans conjecture after ~30 years of remaining open
- ▸The AI identified a graph with fractional flow cost 58, where any practical flow solution costs at least 60
- ▸This demonstrates LLMs' emerging capability in rigorous mathematical reasoning and conjecture-solving
Summary
ChatGPT has discovered a counter-example to the Dinitz–Garg–Goemans (DGG) conjecture, a long-standing open problem in graph theory and combinatorial optimization that remained unsolved for approximately 30 years. Using ChatGPT 5.6 Pro through interactive conversation, researchers identified a specific graph configuration that disproves the conjecture. The counter-example demonstrates a fractional flow cost of 58, while any unsplittable flow with capacity violation ≤15 requires a cost of at least 60, definitively proving the conjecture false.
This breakthrough marks a significant milestone in AI-assisted mathematical discovery, demonstrating that large language models can contribute meaningfully to rigorous mathematical reasoning beyond pattern recognition and text generation. The discovery was made through collaborative dialogue with the AI system, highlighting the potential of LLMs as analytical tools for exploring complex problem spaces in pure mathematics.
- The breakthrough was achieved through interactive conversation with ChatGPT 5.6 Pro
- The finding has implications for combinatorial optimization and flow problems in theoretical computer science
Editorial Opinion
This is a watershed moment for AI in mathematics—not because ChatGPT replaces human mathematicians, but because it augments their toolkit. The discovery validates LLMs as collaborative partners in exploring mathematical frontiers that might otherwise require months or years of human investigation. While questions remain about how much of the breakthrough was machine computation versus the quality of prompting, the result suggests AI systems can accelerate progress on other long-standing open problems across mathematics and theoretical computer science.
