Unverified50% confidenceFactExact time
Soft-MTP decouples the KL self-distillation signal from the noise residual, so that during gradient backpropagation the gradient flows only through the KL self-distillation part.
1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe gradient of standard MTP contains two components: a KL self-distillation signal and a noise residual.80% similarUnverifiedLong skip connections directly short-circuit the diffusion noise to the locations where it is needed, weakening residual noise signals in the backbone network, thereby purifying the backbone and improving feature quality.64% similarUnverifiedThe doubly sparse model simultaneously parametrizes the nonzero support of the signal and the nonzero structure of the noise, examining how the BBP-type phase transition moves with the two types of sparsity.58% similarUnverified过程噪声协方差矩阵Q过小会导致滤波器对机动响应迟钝，Q过大则会放大测量噪声导致跟踪曲线抖动57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1118523API
curl https://kongchang.com/api/v1/knowledge/claims/1118523MCP
get_claim(id=1118523)