Unverified50% confidenceFactExact time
HELM (Stanford) comprehensively evaluates models across multiple dimensions including accuracy, calibration, and robustness.
1
Sources
50%
Confidence
Long-term
Relevance
9/12/2026
First Seen
Sources
Related Entities
Related Claims
Unverified斯坦福的IFEval、清华的FollowBench是评估模型指令遵循能力的基准测试70% similarUnverifiedHELM的7大评估指标包括准确性、校准性、鲁棒性、公平性、偏见与刻板印象、毒性以及效率68% similarUnverifiedThe quality of tool descriptions directly affects the accuracy of model invocation.66% similarUnverifiedA Stanford University research team proposed Citation Precision and Citation Recall metrics to evaluate citation quality.65% similarUnverified校准性（Calibration）指模型表达的置信度与实际准确率之间的一致程度，是评估AI模型质量的核心指标之一64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/911159API
curl https://kongchang.com/api/v1/knowledge/claims/911159MCP
get_claim(id=911159)