[KongchangAI]
Unverified50% confidenceFactExact time

The expression `h_mtp.detach() - h_mtp` is always numerically zero; the hidden state is unchanged during forward propagation, but during backpropagation the gradient flows through the MTP branch in negated form.

1
Sources
50%
Confidence
Long-term
Relevance
10/30/2026
First Seen

Sources

Related Entities

Related Claims

Cite This Claim

Stable URI
https://kongchang.com/claim/1118534
API
curl https://kongchang.com/api/v1/knowledge/claims/1118534
MCP
get_claim(id=1118534)