SOURCE-LINKED INTELLIGENCE

High-probability guarantees for linear accessibility in feature superposition

arXiv · AI, language, vision and robotics · article · Sep 9, 2026 · UTC

Neural networks can leverage feature superposition to encode more concepts than dimensions, but cross-feature interference constrains the linear accessibility of simultaneously active features. By framing linear accessibility as a compressed sensing problem, we derive high-probability bounds for fixed supports under subgaussian noise, proving the sufficient dimension scales linearly ($d=O_{\varepsilon}(k \log m)$) rather than prior worst-case quadratic limits. We then validate these bounds across system parameters through Gaussian-tail approximations. These results quantify the geometric const

Read original source ↗ Open in workspace

recordType

paper

region

Global