Expired85% confidenceFactTime unknown
Top large language models have generally surpassed 90% on MMLU, causing the test's discriminative power to decline.
1
Sources
85%
Confidence
Medium-term (~90 days)
Relevance
5/31/2026
First Seen
Valid until: 8/29/2026(expired)
Sources
Claude 4.5 vs Gemini 3 Pro: A Comprehensive Coding Showdown
bilibiliClaudeEvangelist
Related Entities
Related Claims
UnverifiedSome models have already surpassed 90% of the human average score on MMLU, causing their discriminative power to drop sharply.84% similarUnverifiedOn code capability benchmarks such as HumanEval, top large language models have already achieved pass rates exceeding 90%.75% 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70% similarUnverifiedBy mid-2024, top models had generally exceeded scores of 85–90% on MMLU, causing the benchmark's discriminative power to drop sharply.70% similarUnverified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66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/4056API
curl https://kongchang.com/api/v1/knowledge/claims/4056MCP
get_claim(id=4056)