pith. sign in

Nevertheless, even at8๐‘˜, a commonly used long-context setting, ZEDA still achieves approximately 20% speedup, demonstrating its practical usability

The speedup gradually diminishes as sequence length increases · 2025

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it
browse 1 citing papers

fields

cs.LG 1

years

2026 1

verdicts

UNVERDICTED 1

clear filters

representative citing papers

Post-Trained MoE Can Skip Half Experts via Self-Distillation

cs.LG ยท 2026-05-18 ยท unverdicted ยท novelty 6.0

ZEDA turns post-trained static MoE models into dynamic ones via zero-output expert injection and two-stage self-distillation, cutting over 50% expert FLOPs on Qwen3-30B-A3B and GLM-4.7-Flash with small accuracy drops across 11 benchmarks.

citing papers explorer

Showing 1 of 1 citing paper after filters.

  • Post-Trained MoE Can Skip Half Experts via Self-Distillation cs.LG ยท 2026-05-18 ยท unverdicted ยท none ยท ref 43

    ZEDA turns post-trained static MoE models into dynamic ones via zero-output expert injection and two-stage self-distillation, cutting over 50% expert FLOPs on Qwen3-30B-A3B and GLM-4.7-Flash with small accuracy drops across 11 benchmarks.