ResearchBenchmarksCode Generation on HumanEval (Pass@1, GSS@10)Follow58.66Pass@1GRPO-MA55.997656.688857.3858.0712Sep 29, 2025Evaluation ResultsMethodMethodLinksPass@1GSS@10GRPO-MATN=4, AN=4, S/S=93.45TN=4, AN=4, S/S=93.452025.0958.6610GRPOTN=16, AN=1, S/S=186.91TN=16, AN=1, S/S=186.912025.0957.825GRPOTN=8, AN=1, S/S=104.83TN=8, AN=1, S/S=104.832025.0956.8324GRPOTN=4, AN=1, S/S=76.21TN=4, AN=1, S/S=76.212025.0956.16