ResearchBenchmarksMathematical Reasoning on AIME 2024-2026 & HMMT 2025 SuiteFollow1.88Accuracy ChangeTRS-1.2296-0.42230.3851.1923Apr 23, 2026Evaluation ResultsMethodMethodLinksAccuracy ChangeOutput Token Change (%)Cost Change (%)TRSModel=Doubao-1.8Model=Doubao-1.82026.041.884.92.8TRSModel=Gemini-3-FlashModel=Gemini-3-Flash2026.040.928.69TRSModel=GPT-OSS-120BModel=GPT-OSS-120B2026.040.58.16.5TRSModel=GPT-OSS-20BModel=GPT-OSS-20B2026.04-0.65.34.3TRSModel=Doubao-2.0-ProModel=Doubao-2.0-Pro2026.04-1.1117.515.7