Apex Intelligence Releases First Evaluation Results of Automated AI Research System, Setting Three SOTA Records

First evaluation results, released on September 9, for an automated AI research system from Apex Intelligence (超衍智能) show the system autonomously identifying improvement points, running experiments, learning from evidence, and feeding those lessons into the next cycle, according to the company. The system set new state-of-the-art scores on three benchmarks and beat automated systems from Stanford, NVIDIA, and Recursive in direct comparisons. The release did not name the benchmarks or report margins.

Original sources (Chinese)