清华学霸与AI比做高考压轴题，谁会赢？

Group 1 - The competition between Tsinghua Yao Class students and AI showcased significant advancements in AI's reasoning and problem-solving capabilities, particularly in high-stakes academic scenarios [2][4] - Tsinghua Yao Class students completed the exam questions in 10 minutes with only one incorrect answer, while the AI provided correct solutions shortly after input confirmation [2][4] - The AI's reasoning process was noted to be clear and aligned with the students' thought processes, enhancing understanding of problem-solving methods [2] Group 2 - AI models have shown substantial improvement in their reasoning abilities, with DeepSeek-R1 being highlighted for its enhanced performance in educational contexts [4] - Previous assessments indicated that last year's AI models struggled with science subjects, but recent tests have shown models achieving scores comparable to top-tier universities [4] - OpenAI's o3-mini demonstrated high reasoning capabilities, solving over 32% of problems in the FrontierMath benchmark, which is a significant achievement in mathematical problem-solving [4] Group 3 - Google introduced AlphaProof and AlphaGeometry 2, which successfully solved four out of six problems from the 2024 International Mathematical Olympiad, indicating a leap in AI's mathematical reasoning [5] - Alibaba's Qwen3 model achieved a score of 81.5 in the AIME25 assessment, setting a new record for open-source models [6] - The AI contribution to the online education market is projected to increase from 7% to 16% between 2023 and 2027, reflecting the growing integration of AI in educational settings [6]