Unverified50% confidenceBenchmarkExact time
在基于推理的问答任务上，最强的闭源模型准确率超过90%
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/11/2026
First Seen
Valid until: 12/10/2026
Sources
Related Entities
Related Claims
Unverified当前顶级模型的单步准确率约在90%至95%之间77% similarUnverified实验数据显示关键信息位于输入序列首尾时模型检索准确率可高达90%以上，位于中段时可能骤降至60%以下77% similarVerified完美校准的模型在所有声称70%置信度的预测中准确率恰好为70%73% 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73% similarVerifiedSpeculative Decoding在草稿模型预测准确率达到60-80%时，整体推理速度可提升2-3倍72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/902745API
curl https://kongchang.com/api/v1/knowledge/claims/902745MCP
get_claim(id=902745)