Unverified50% confidenceBenchmarkExact time
On six new International Mathematical Olympiad problems, all models with room for improvement scored higher than their own providers' coding agents under the AutoFyn framework.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/11/2026
First Seen
Valid until: 12/10/2026
Sources
Related Entities
Related Claims
Unverified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% similarUnverifiedImageNet竞赛曾暴露研究者通过反复调优超参数刷榜、导致模型基准表现与实际部署效果出现显著差距的问题60% similarUnverified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60% similarUnverifiedLarge language models are prone to hallucinations when handling complex and ambiguous requirements; the larger and more ambiguous the scope, the higher the probability of AI errors, so reducing scope for MVP lowers the risk of AI mistakes.60% similarUnverified强化学习领域长期存在可复现性问题，即使使用相同算法和超参数，不同代码实现之间的性能差异可高达数倍60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/893611API
curl https://kongchang.com/api/v1/knowledge/claims/893611MCP
get_claim(id=893611)