SOURCE-LINKED INTELLIGENCE

Output Dilution: Redundant but Fragile Representations in MoE Models

arXiv · AI, language, vision and robotics · article · Aug 25, 2026 · UTC

Mixture-of-Experts (MoE) models appear to encode moral content as robustly as dense models, yet prove far more fragile in their encoding. In OLMoE-1B-7B, linear probes recover moral valence from nearly every expert-layer combination, with mean peak-layer accuracy above 90%. But these representations collapse under levels of activation noise that a dense model of matched size easily tolerates, with a 4.2-fold difference in robustness. We trace this to output dilution. Because the MoE block averages across active experts before contributing to the residual stream, the feedforward signal reaching

Read original source ↗ Open in workspace

recordType

paper

region

Global