CyberIntel ⬡ News
★ Saved ◆ Cyber Reads
← Back ◬ AI & Machine Learning Aug 13, 2026

Detecting a Route Flip Is Easier Than Knowing Whether to Fix It: Causal Route-Mediated Damage in Quantized Mixture-of-Experts

arXiv AI Archived Aug 13, 2026 ! Full text unavailable

arXiv:2608.11212v1 Announce Type: new Abstract: Top-k Mixture-of-Experts (MoE) routing is discontinuous, so a deployment-motivated numerical disturbance -- simulated 4-bit KV-cache quantization read by a protected BF16 gate -- pushes tokens across decision boundaries and flips which experts fire. This paper proposes no new mitigation; it supplies a causal apparatus, empirical findings, and a detection-limit result. A four-run apparatus prices the route-mediated fraction (RMF) of quantization dam

Full text unavailable — view original
✦ AI Summary · Claude Sonnet


    Full text unavailable.
    Open original ↗
    💬 Team Notes
    Article Info
    Source
    arXiv AI
    Category
    ◬ AI & Machine Learning
    Published
    Aug 13, 2026
    Archived
    Aug 13, 2026
    Full Text
    ✗ Not available
    Open Original ↗