Unverified50% confidenceFactExact time

MMLU、TruthfulQA等主流基准测试已开始关注弃权维度，但业界尚未形成统一的弃权评分标准

1

Sources

50%

Confidence

Medium-term (~90 days)

Relevance

9/28/2026

First Seen

Valid until: 12/27/2026

Sources

一句"不要猜测"让AI幻觉从71%降至20%

Related Entities

MMLUproduct TruthfulQAproduct

Related Claims

Unverified目前尚未出现像MMLU那样被普遍引用的统一LLM安全评测标准71% similar Unverified目前业界尚无公认的Prompt质量评估标准71% similar Unverified当前VLA领域缺乏可复现、可横向对比的标准化评测基准，多数VLA论文在各自定制的实验环境中报告结果67% similar Unverified该求解声明的真实性、是否经过同行评审、是否获得克雷数学研究所认可均缺乏独立佐证66% similar Unverified目前学界尚无公认的标准来衡量一个模型是否「足够安全」，模型评估（Model Evaluation）领域尚不成熟64% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/955815

API

curl https://kongchang.com/api/v1/knowledge/claims/955815

MCP

get_claim(id=955815)