Unverified50% confidenceFactExact time
A large-scale study published by Bastani et al. in PNAS 2025 showed that students who practiced math with AI directly giving the answers performed worse in later exams than students with no AI assistance at all.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedMultiple independent studies in 2023 showed that some top models scored extremely high on standard mathematical reasoning tests but performed significantly worse on equivalent rephrased versions, indicating that the models learned test patterns rather than genuine reasoning ability.70% similarUnverified使用AI辅助的学生在不能使用AI的闭卷考试中成绩下降了20%67% similarUnverifiedAndrew Ng believes that AI makes fewer mistakes than most people imagine, especially when prompted correctly, and that models in 2022 and 2023 made far more errors than they do now.65% similarUnverifiedFrontier Math 是由 EpochAI 创建的数学基准测试,此前 AI 模型在该测试上的通过率不足 2%63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1006102API
curl https://kongchang.com/api/v1/knowledge/claims/1006102MCP
get_claim(id=1006102)