Unverified50% confidenceDecision RuleExact time
The 95% equivalence threshold means: when sampled evaluation shows that the proportion of outputs where the lightweight model is equivalent to the flagship model falls below 95%, the gateway automatically switches all traffic back to the flagship model.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/21/2026
First Seen
Valid until: 1/19/2027
Sources
Related Entities
Related Claims
UnverifiedDuring testing, cases emerged where a model claimed 95% confidence in a move but scored 0 points according to the evaluation criteria.72% similarUnverified参数量在1亿以下的轻量模型经领域数据微调后可达95%以上意图分类准确率，单次推理延迟控制在5-20ms以内69% similarUnverified当所有模型在某评测上都能拿90分时该评测就应退场，即基准饱和现象69% similarUnverified校准的不确定性区间意味着：若模型声称 90% 置信区间为某范围，则在大量测试样本中约 90% 的真实值应落入该范围。66% similarUnverifiedThe throughput and labor costs of a HITL system can be estimated from the confidence distribution; if 80% of samples exceed the confidence threshold, theoretically only the remaining 20% need human processing.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1080695API
curl https://kongchang.com/api/v1/knowledge/claims/1080695MCP
get_claim(id=1080695)