GPT-5.6 trains 64 agents and provides proof of 50 years of conjectures within an hour.

robot
Abstract generation in progress
According to Beating Monitor’s observation, OpenAI said that GPT-5.6 Sol Ultra calls 64 sub-agents, using less than 1 hour, to provide a proof for the proposed “loop double-covering conjecture” from about 50 years ago.

OpenAI has released the full prompt and the entire proof text, and open-sourced a Lean formalized version. The relevant code has passed Lean kernel checks. However, this is still a result published by OpenAI. The mathematical community has not yet completed an independent review, so for now it can only be said to have “provided a proof,” not that it has “officially solved” it.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned