Unverified50% confidenceFactExact time
GEC任务的核心评估指标F0.5更看重精确率而非召回率
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedGLM5.1相比GLM5在编程能力（Coding Evaluation）测评数据上有显著提升66% similarUnverified稳健的具身AI评测基准应考虑样本效率和跨随机场景的泛化能力，而不仅是任务完成率和空间精度66% similarUnverified五项测试的综合结论显示，提示词的精确程度是决定AI蓝图生成质量的核心变量，模糊指令导致不符合预期的结果，具体指令则能获得接近完美的结果64% similarUnverified对抗 RL 的鲁棒性评估应在目标部署场景中重新验证，而非盲目照搬既有结论64% similarUnverifiedFDE入行的六条硬标准衡量的核心是亲手做的项目是否够格算一次真实交付63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/907886API
curl https://kongchang.com/api/v1/knowledge/claims/907886MCP
get_claim(id=907886)