Image Understanding on COCO (Score)
69.3ScoreTop-k Routing
Evaluation Results
| Method | Links | |
|---|---|---|
| Top-k RoutingExpert skipping ratio (ρ)=0, Number of experts (k)=8, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 69.3 | |
| MoDESExpert skipping ratio (ρ)=0.6, Number of experts (k)=3, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 68.22 | |
| MoDESExpert skipping ratio (ρ)=0.73, Number of experts (k)=2, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 67.89 | |
| DiEPExpert skipping ratio (ρ)=0.6, Number of experts (k)=3, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 67.17 | |
| MC-MoEExpert skipping ratio (ρ)=0.6, Number of experts (k)=3, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 66.87 | |
| NAEEExpert skipping ratio (ρ)=0.6, Number of experts (k)=3, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 66.05 | |
| Top-k RoutingExpert skipping ratio (ρ)=0.6, Number of experts (k)=3, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 65.47 | |
| MoDESExpert skipping ratio (ρ)=0.85, Number of experts (k)=1, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 65.16 | |
| MC-MoEExpert skipping ratio (ρ)=0.73, Number of experts (k)=2, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 64.24 | |
| NAEEExpert skipping ratio (ρ)=0.73, Number of experts (k)=2, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 63.31 | |
| DiEPExpert skipping ratio (ρ)=0.73, Number of experts (k)=2, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 63.1 | |
| MC-MoEExpert skipping ratio (ρ)=0.85, Number of experts (k)=1, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 60.12 | |
| Top-k RoutingExpert skipping ratio (ρ)=0.73, Number of experts (k)=2, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 58.28 | |
| DiEPExpert skipping ratio (ρ)=0.85, Number of experts (k)=1, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 55.78 | |
| NAEEExpert skipping ratio (ρ)=0.85, Number of experts (k)=1, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 50.64 | |
| Top-k RoutingExpert skipping ratio (ρ)=0.85, Number of experts (k)=1, Backbone=InternVL-3.5-30B-A3B-HF2025.11 | 26.01 |