ResearchBenchmarksClosed Loop RLHF on HH 0% noiseFollow61.6Win Rateα-URM54.73656.51858.360.082Jul 3, 2026Evaluation ResultsMethodMethodLinksWin Rateα-URMBase Model=Qwen-3-1.7B...Base Model=Qwen-3-1.7B, RL Algorithm=GRPO, Number of generations=2, Epochs=32026.0761.6URMBase Model=Qwen-3-1.7B...Base Model=Qwen-3-1.7B, RL Algorithm=GRPO, Number of generations=2, Epochs=32026.0760.3RMBase Model=Qwen-3-1.7B...Base Model=Qwen-3-1.7B, RL Algorithm=GRPO, Number of generations=2, Epochs=32026.0755