Unverified50% confidenceFactExact time
在该任务中，找出问题所在位置（定位F1普遍在0.78以上）比准确判断问题类型（标签F1约0.5）容易得多
1
Sources
50%
Confidence
Long-term
Relevance
9/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedQwen 3.6 Max中性模式下歧义检测F1仅16%，平均每任务只问0.07个问题，但一旦提问澄清准确率高达94.7%68% similarUnverified当每个目标的雷达检测点从1个增加到5个时，Macro F1从0.381跃升至0.76464% similarUnverifiedCriticGen在基于标准的理由和可执行建议上的F1分数分别从0.6369/0.5994提升到0.7554/0.790063% similarUnverifiedV4 Pro Max满血推理模式下算法竞赛3206分、编程93.5分、科学问答90.1分、软件工程80.6分、终端操作67.9分60% similarUnverifiedF1分数是召回率和精确率的调和平均数59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/919722API
curl https://kongchang.com/api/v1/knowledge/claims/919722MCP
get_claim(id=919722)