Unverified50% confidenceFactExact time
When the data distribution is supported on a d0-dimensional smooth submanifold, the Wasserstein-p dimension d_p* is typically of the same order as d0, which is much smaller than the ambient dimension D.
1
Sources
50%
Confidence
Long-term
Relevance
10/5/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedMLA compresses the KV Cache into low-dimensional latent vectors via low-rank decomposition, with d_latent approximately 1/8 of d_model, reducing KV Cache VRAM usage to 5%–13% of traditional multi-head attention.55% similarUnverified对于一个100万面的模型，LOD1可简化至25万面、LOD2简化至6万面，远距离渲染时GPU负载可降低75%以上55% similarUnverifiedQuantized DeepSeek models require several GB of VRAM, and Stable Diffusion inference typically requires 4–8 GB of VRAM.54% similarUnverifiedllama.cpp通过INT4/INT8量化将模型权重压缩至原始大小的25%以下53% similarUnverifiedStable Diffusion XL仅模型参数就占用约6.5GB显存52% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/976007API
curl https://kongchang.com/api/v1/knowledge/claims/976007MCP
get_claim(id=976007)