Unverified50% confidenceFactExact time
In multi-reference image tasks, each additional reference image increases the context length during the attention computation stage, and VRAM usage and computational load tend to rise nonlinearly.
1
Sources
50%
Confidence
Long-term
Relevance
9/28/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWhen resolution doubles, the attention computation the model must handle often grows fourfold or more, causing elapsed time to increase non-linearly.78% similarUnverifiedStandard attention's computation grows quadratically with sequence length, whereas the low-rank bottleneck decomposes the transformation into two steps: first compressing to r dimensions and then expanding.77% similarUnverified在推理阶段内存带宽往往比纯算力更早成为瓶颈，注意力机制需在每步生成时读取全部KV缓存，上下文长度增大时内存读写开销呈平方级增长77% similarUnverifiedThe attention score of key evidence has an upper bound, while the maximum score among all effective distractors grows as the number of distractors increases.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/956082API
curl https://kongchang.com/api/v1/knowledge/claims/956082MCP
get_claim(id=956082)