Unverified50% confidenceFactExact time
GTR-Depth uses no softmax attention anywhere in the entire network.
1
Sources
50%
Confidence
Long-term
Relevance
10/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedGTR-Depth uses the same backbone network as its predecessor, containing 12 gated linear attention blocks.71% similarUnverifiedDeepSeek-V3.2-Exp首次引入自研的DeepSeek Sparse Attention（DSA）稀疏注意力技术63% similarUnverifiedGTR-Depth is a continuation of the GTR series of models, focused on the depth estimation task.58% similarUnverifiedIn DeepSeek-V4 optimizations, skipping unnecessary topk/router during the decoding phase brings a 3.4% end-to-end TTFT improvement, and workspace reuse brings a 3.9% TTFT improvement.57% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/971170API
curl https://kongchang.com/api/v1/knowledge/claims/971170MCP
get_claim(id=971170)