ResearchBenchmarksMath Reasoning on AIME 25 (Avg@32)Follow24.12Avg@32TIC-GRPO4.55769.636314.71519.7937Aug 4, 2025Evaluation ResultsMethodMethodLinksAvg@32TIC-GRPOModel=Qwen3-8BModel=Qwen3-8B2025.0824.12GRPOModel=Qwen3-8BModel=Qwen3-8B2025.0822.9GSPOModel=Qwen3-8BModel=Qwen3-8B2025.0822.5TIC-GRPOModel=Qwen3-1.7BModel=Qwen3-1.7B2025.086.98GSPOModel=Qwen3-1.7BModel=Qwen3-1.7B2025.086.24GRPOModel=Qwen3-1.7BModel=Qwen3-1.7B2025.085.31