AI Achieves Major Breakthrough in Mathematics
Key Questions
What major math problem did GPT-5.6 Sol Ultra solve?
GPT-5.6 Sol Ultra proved the Cycle Double Cover Conjecture, a 50-year-old unsolved problem from Wikipedia's list. It used 64 subagents to generate the proof in under an hour, marking the first LLM-solved entry from that list.
How does GPT-5.6 Sol Ultra's achievement compare to earlier AI math work?
Unlike prior efforts like the Erdős Unit Distance problem, this marks a novel proof of an open conjecture. Grok 4.5 had earlier found a counterexample to hypercontractivity on the 4-sphere.
What supporting tools and benchmarks are advancing LLM math capabilities?
The LeanDojo ecosystem is maturing, supported by the new ITPEval benchmark and four Lean-related papers presented at ICML 2026 workshops.
Why is this considered a breakthrough in AI mathematics?
It demonstrates an LLM independently solving a long-standing open problem from a public list of unsolved conjectures. The rapid generation via multi-agent setup highlights scaling progress in formal reasoning.
Is the proof publicly available?
Yes, a paper titled 'GPT-5.6 Sol Ultra produces proof of the Cycle Double Cover Conjecture' has been released as a PDF detailing the result.
GPT-5.6 Sol Ultra proved the Cycle Double Cover Conjecture using 64 subagents in under an hour—first LLM-solved problem from Wikipedia's list of unsolved problems. Grok 4.5 earlier constructed a counterexample to hypercontractivity on the 4-sphere. LeanDojo ecosystem maturing with ITPEval benchmark and four Lean-related papers at ICML 2026 workshops.