ResearchBenchmarksMathematical Reasoning on AIME, HMMT, BRUMO average '25Follow48.65Pass@4XRPO39.841242.128144.41546.7019Oct 8, 2025Evaluation ResultsMethodMethodLinksPass@4Pass@8Pass@16Pass@32XRPOBackbone=Qwen3-1.7BBackbone=Qwen3-1.7B2025.1048.6555.9262.0765.56GSPOBackbone=Qwen3-1.7BBackbone=Qwen3-1.7B2025.1046.0952.0157.662.22DAPOBackbone=Qwen3-1.7BBackbone=Qwen3-1.7B2025.1042.3747.552.1456.67TPO-SBackbone=Qwen3-1.7BBackbone=Qwen3-1.7B2025.1040.1847.1654.6963.33