Unverified50% confidenceFactExact time
The beta value of the F2 score is set to 2, meaning recall is weighted 4 times more than precision
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified金融风控中可以使用F2分数(β=2)来更强调查全率，垃圾邮件过滤中可以使用F0.5分数(β=0.5)来更强调查准率68% similarUnverified4-bit quantization compared to FP16 typically results in a 0.5–2% drop in benchmark accuracy.64% similarVerifiedThe computational load of the attention mechanism is proportional to the square of the sequence length, meaning doubling the context increases computation fourfold.63% similarUnverifiedThe intermediate results of FFT are complex numbers (real part + imaginary part), causing memory usage to double.63% similarUnverifiedIn a binary classification confusion matrix, the F1 score is the harmonic mean of precision and recall.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/893422API
curl https://kongchang.com/api/v1/knowledge/claims/893422MCP
get_claim(id=893422)