Unverified50% confidenceFactExact time
业界常用的AI评测基准包括MMLU、HumanEval、GSM8K等，每个基准只能衡量特定维度的能力
1
Sources
50%
Confidence
Long-term
Relevance
8/26/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAI模型能力评估长期依赖MMLU、HumanEval、GSM8K等标准化基准集77% similarUnverifiedAI领域常见评测集包括MMLU、HumanEval、MATH、GPQA，均有公开标准题库和评分方法可独立复现75% similarUnverifiedMMLU、HumanEval、GSM8K等评测榜单曾是AI模型竞争与公众注意力的焦点73% similarUnverifiedMLflow、LangSmith、Weights & Biases等工具链用于解决AI生产环境的可观测性与评估挑战67% similarUnverifiedAI Agent评估平台包括Braintrust、LangSmith、Weights & Biases，提供从数据集管理、评估执行到结果可视化的完整工作流67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/804697API
curl https://kongchang.com/api/v1/knowledge/claims/804697MCP
get_claim(id=804697)