InfoMamba：一种无注意力的混合Mamba-Transformer模型

The challenge of balancing fine-grained local modeling with long-range dependency capture under computational constraints is a central issue in sequence modeling. While Transformers excel in token mixing, they suffer from quadratic complexity, whereas Mamba-style selective state-space models (SSMs) scale linearly but often struggle to capture high-rank and synchronous global interactions. A consistency boundary analysis is presented that characterizes when diagonal short-memory SSMs can approximate causal attention and identifies existing structural gaps. Motivated by this analysis, InfoMamba is proposed as an attention-free hybrid architecture. InfoMamba replaces token-level self-attention with a concept bottleneck linear filtering layer, enhancing the model's efficiency and performance.