Unverified50% confidenceFactExact time
Group GEMM is the bottleneck of MoE computation, as the differing numbers of tokens received by different experts require simultaneously executing a batch of small matrix multiplications with varying shapes.
1
Sources
50%
Confidence
Long-term
Relevance
10/4/2026
First Seen
Sources
Related Entities
Related Claims
Unverified强行取极值样本会引入离群点，迫使SVM构建更复杂的决策边界，高阶poly核在这种数据上收敛更慢63% similarUnverifiedIn large-scale GPU matrix operations, the associative law of floating-point addition does not strictly hold, and different parallel reduction orders, batch sizes, and memory scheduling can cause minor discrepancies in logits.61% similarUnverifiedClassical sieve methods have a fundamental bottleneck known as the parity barrier, which prevents distinguishing between integers with an odd number and those with an even number of prime factors.61% similarUnverifiedFlattening a matrix and feeding it to an MLP results in a non-fixed input dimension due to variable row counts, and breaks permutation invariance.61% similarUnverifiedSol的论证使用堆积论证：一旦单零点密度超过1/2就会强制出现线性数量级的紧密三元组，每个三元组对应一个正谱缺陷60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/972876API
curl https://kongchang.com/api/v1/knowledge/claims/972876MCP
get_claim(id=972876)