AI agents collaborate to solve the 11D kissing number problem, pushing the lower bound to 604

ME News reports that on April 14 (UTC+8), according to monitoring by 1M AI News, AI infrastructure company Together AI has launched EinsteinArena, a platform enabling AI agents to collaboratively compete on open mathematical problems. Agents can query problems via API, submit solutions, and exchange intermediate results in discussion forums, with leaderboards updated in real time. Since its launch on March 19, the platform has achieved new optimal solutions on 11 open mathematical problems as of April 11. The most notable breakthrough concerns the 11-dimensional Kissing Number problem—determining the maximum number of equal-sized spheres that can simultaneously touch a central sphere in high-dimensional space. The previous best lower bound of 592 was set by Google DeepMind’s AlphaEvolve, raised to 593. On EinsteinArena, AI agents directly pushed this lower bound to 604. The breakthrough itself exemplifies multi-agent collaboration: on April 8, one agent submitted a promising construction with minor sphere overlaps; over the next 48 hours, multiple agents iteratively refined it, ultimately confirming validity through high-precision verification (increasing precision from standard floating-point to 1e-50 levels). Competition on the platform is intense: the Erdős Minimum Overlap problem attracted over 10 agents submitting 22 distinct constructions and initiating 37 discussion threads; the Second Autocorrelation Inequality problem saw 17 agents submit 18 solutions, with a student collaborating with Claude discovering a new lower bound—only to be surpassed shortly after by another agent while the student was writing a blog post. The platform is fully open-source and supports integration by any AI agent. (Source: BlockBeats)