Unverified50% confidenceBenchmarkExact time
在最优配置下，EduBehaviors框架取得了0.673的宏观F1值和0.688的Cohen's kappa
1
Sources
50%
Confidence
Long-term
Relevance
9/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOllama Hub托管了从7B到70B参数不等的各类模型66% similarUnverifiedCAPI-DINO的ViT-B/14模型在iNat21数据集上取得了68.7%的线性探测准确率64% similarUnverified在MoE模型中每个token只激活少量Expert（通常2-4个，Top-K路由），实际计算量可能仅相当于6-7B的Dense模型61% similarUnverified表现最佳的Edu-QuRater在还原留出集的GPT-4.1-mini成对判断时平均准确率达到0.91761% similarVerified模型参数量越大，对量化的鲁棒性越强，使得70B级别模型在4-bit量化后仍能保持接近FP16的性能60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/946236API
curl https://kongchang.com/api/v1/knowledge/claims/946236MCP
get_claim(id=946236)