Unverified50% confidenceFactExact time
主流基准测试如MMLU、HumanEval、SWE-bench通常仅覆盖数千至数万个标准化问题，存在场景覆盖有限和训练数据污染的潜在风险
1
Sources
50%
Confidence
Long-term
Relevance
7/19/2026
First Seen
Sources
Related Claims
UnverifiedOfficial benchmarks such as MMLU and HumanEval carry risks of test set leakage and targeted optimization.81% similarUnverifiedCurrent AI model evaluation relies primarily on standardized benchmark tests (such as MMLU, HumanEval, and GSM8K), but these face risks of data contamination and disconnection from real-world applications.74% similarUnverifiedStandard AI benchmarks include MMLU (a comprehensive knowledge test covering 57 subjects), GPQA (graduate-level professional question answering), SWE-bench (real-world software engineering tasks), HumanEval and MBPP (programming ability), and MATH and GSM8K (mathematical reasoning)72% similarUnverified当前主流基准测试体系如HumanEval、MBPP等代码评估集往往关注平均性能，对分布式的场景特定退化现象敏感度不足71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/558192API
curl https://kongchang.com/api/v1/knowledge/claims/558192MCP
get_claim(id=558192)