Unverified50% confidenceBenchmarkExact time
Wolf Defender主模型的真实场景良性特异度从v1的66.85%提升至v2的96.63%
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
8/31/2026
First Seen
Valid until: 11/29/2026
Sources
Related Entities
Related Claims
UnverifiedWolf Defender v2的攻击检测能力基本保持原有水平，Qualifire F1为95.14%，Jayavibhav F1为97.84%76% similarUnverifiedWolf Defender v2结合短样本（256 token）与完整长窗口（2048 token）进行混合训练62% similarUnverifiedEfficientNet 从 B0 到 B7，模型参数从 5.3M 增长到 66M，Top-1 精度从 77.1% 提升到 84.3%55% similarUnverifiedbuffalo_l在MS1MV3数据集（约300万张图像）上训练，LFW准确率接近99.8%54% similarUnverifiedNVFP4量化方案相比BF16基线实现了1.5倍的推理加速，同时在top-1准确率上保留了92%到97%的水平53% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/827598API
curl https://kongchang.com/api/v1/knowledge/claims/827598MCP
get_claim(id=827598)