Verified75% confidenceFactExact time
Some models have already surpassed 90% of the human average score on MMLU, causing their discriminative power to drop sharply.
3
Sources
75%
Confidence
Medium-term (~90 days)
Relevance
7/24/2026
First Seen
Valid until: 10/22/2026
Sources
多智能体与代码生成实测：新一代大模型能力深度解析
Related Claims
Unverified基准污染（Benchmark Contamination）导致HumanEval、MMLU等主流基准区分度下降，部分模型HumanEval得分超过95%、MMLU超过90%但实际能力平庸80% similarVerifiedBy mid-2024, top models had generally exceeded scores of 85–90% on MMLU, causing the benchmark's discriminative power to drop sharply.80% similarUnverified一项针对GPT-4的研究显示，对MMLU题目进行语义等价改写后，模型准确率在部分子集上下降超过10个百分点73% similarUnverifiedThe pass rate of top models in 2025 on the HumanEval benchmark has exceeded 90%.71% similarUnverifiedWhen multiple models all achieve accuracy rates above 90% on benchmarks such as MMLU and HumanEval, user attention shifts to practical metrics such as inference latency, service availability, and total cost of ownership.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/607159API
curl https://kongchang.com/api/v1/knowledge/claims/607159MCP
get_claim(id=607159)