ResearchBenchmarksMathematical Reasoning on AIME 2025 (Acceptance length, Delta %)Follow4.9Acceptance LengthTTS2.823.363.94.44May 10, 2026Evaluation ResultsMethodMethodLinksAcceptance LengthDelta (%)TTSTarget Model=Qwen/Qwen...Target Model=Qwen/Qwen3.5-122B2026.054.971.8TTSTarget Model=Qwen/Qwen...Target Model=Qwen/Qwen3-4B2026.054.747TTSTarget Model=Qwen/Qwen...Target Model=Qwen/Qwen3-8B2026.054.233.1DFlashTarget Model=Qwen/Qwen...Target Model=Qwen/Qwen3-4B2026.053.2—DFlashTarget Model=Qwen/Qwen...Target Model=Qwen/Qwen3-8B2026.053.1—DFlashTarget Model=Qwen/Qwen...Target Model=Qwen/Qwen3.5-122B2026.052.9—