Unverified50% confidenceReleaseExact time
The paper "AttSVD: Prompt-Adaptive Low-Rank KV Cache Compression via Attention-Guided SVD" was published on arXiv.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/7/2026
First Seen
Valid until: 1/5/2027
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.70% similarUnverifiedMLA 将所有头的 Key 和 Value 联合压缩为低维潜在向量缓存，本质上是对 KV Cache 做低秩分解68% similarUnverifiedGQA groups attention heads to share KV projections, which can compress the KV cache to 1/4 or less of the original size with almost no loss in model quality.67% similarUnverifiedMLA（Multi-head Latent Attention）通过低秩压缩将 KV 缓存的显存占用降低至传统 MHA 的约 1/867% similarUnverifiedKV cache runtime provides a low-cost, fine-grained, real-time intervention capability that fills the gap between model fine-tuning and external scaffolding.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/983253API
curl https://kongchang.com/api/v1/knowledge/claims/983253MCP
get_claim(id=983253)