Where Prefill Expert Reduction Hurts MoE Models: A Few Prompt Tokens Carry Most of the Short-Answer F1 Loss
arXiv:2610.22454v1 Announce Type: new Abstract: Running prompt tokens through fewer routed experts during prefill is a simple way to make mixture-of-experts (MoE) serving cheaper, yet it is unclear what this damages. We locate the damage with token-, layer- and attention-path interventions on short…
Read the full story at arXiv cs.CL ↗
Timeline · 1 report
- 2026-10-19 04:00 · arXiv cs.CL
Where Prefill Expert Reduction Hurts MoE Models: A Few Prompt Tokens Carry Most of the Short-Answer F1 Loss