Experiments show that Parallel-R1 not only brings an average accuracy improvement of up to 8.4% across multiple mathematical ...