Unverified50% confidenceFactExact time
大模型能力边界呈现高度不规则的锯齿状，可能轻松通过美国医师执照考试，却在'9.11和9.9哪个更大'这类小学数学题上出错
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
Unverified引导过程若缺乏严格的数学保证，可能引入难以察觉的系统性误差67% similarUnverified大语言模型在数学证明、代码生成等任务上超越人类平均水平，却在空间推理、计数等简单任务上出现失误，呈现非均匀能力分布65% similarUnverified随着模型规模从百万参数扩展至千亿参数，不理解底层数学的工程师越来越难以诊断训练不稳定、梯度爆炸或模式崩塌等问题64% similarUnverifiedThe TruthfulQA benchmark found that larger models sometimes produce more confident errors, a phenomenon known as the scaling paradox.64% similarUnverified对于数学计算这类对精度要求极高的场景，浮点数溢出、数值不稳定等问题尤为致命63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/579718API
curl https://kongchang.com/api/v1/knowledge/claims/579718MCP
get_claim(id=579718)