LongCat-Flash-Lite's caching strategy: Can N-gram Cache generalize to other models?

LongCat-Flash-Lite Interpretation: A New Efficiency Path for Sparse MoE with N-gram Embeddings

1. Abstract LongCat-Flash-Lite is an open-source large model targeting high-sparsity MoE scenarios: the total parameters are 68.5B, but only about 2.9...