arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2602.10416cs.LGcs.AIcs.CL
收藏

人工智能算法

AI-rithmetic

Google(谷歌)

Alex Bie, Travis Dick, Alex Kulesza, Prabhakar Raghavan, Vinod Raman, Sergei Vassilvitskii

摘要:

现代人工智能系统已成功部署,在国际数学竞赛中赢得奖牌,协助研究工作流程,并证明新的技术引理。然而,尽管他们在高级数学水平上取得了进步,但他们在基本算术上仍然很糟糕,在两个数字相加的简单任务上总是失败。我们对这一现象进行了系统的研究。我们的经验表明,所有的前沿模型遭受显着下降的精度为整数加法作为数字的数量增加。此外,我们表明,这些模型所产生的大多数错误是高度可解释的,可以归因于操作数不对齐或未能正确进行;这两个错误类别分别解释了克劳德Opus 4.1,GPT-5和Gemini 2.5 Pro错误的87.9%,62.9%和92.4%。最后,我们表明,错位错误往往与标记化,并进行错误出现在很大程度上作为独立的随机故障。

英文摘要:

Modern AI systems have been successfully deployed to win medals at international math competitions, assist with research workflows, and prove novel technical lemmas. However, despite their progress at advanced levels of mathematics, they remain stubbornly bad at basic arithmetic, consistently failing on the simple task of adding two numbers. We present a systematic investigation of this phenomenon. We demonstrate empirically that all frontier models suffer significantly degraded accuracy for integer addition as the number of digits increases. Furthermore, we show that most errors made by these models are highly interpretable and can be attributed to either operand misalignment or a failure to correctly carry; these two error classes explain 87.9%, 62.9%, and 92.4% of Claude Opus 4.1, GPT-5, and Gemini 2.5 Pro errors, respectively. Finally, we show that misalignment errors are frequently related to tokenization, and that carrying errors appear largely as independent random failures.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑